# microManager: full documentation > microManager is a free desktop app for macOS, Windows and Linux that runs several coding agents at once (Claude Code, Codex, Grok Build, OpenCode, Gemini CLI and any ACP agent), each in its own git worktree, so you can review their diffs and ship pull requests from one window. This file holds the download page, every microManager docs, feature and agent page, and the guides for switching from Conductor, each as Markdown with its own front matter. The index is at https://micromanager.win/llms.txt. --- title: "Download microManager for Mac, Windows and Linux" description: "Download microManager free for macOS (Apple Silicon, Intel), Windows x64 and Linux x64 (AppImage, .deb), or install it with one verified command." url: "https://micromanager.win/install" updated: "2026-11-09" --- # Download microManager microManager is a free download for macOS (Apple Silicon and Intel), Windows (x64) and Linux (x64, as an AppImage or a .deb), and needs no account. One command can install it instead, in Terminal on macOS and Linux or in PowerShell on Windows, checking the download’s SHA-512 first. Updated 9 November 2026 [Download for this computer](https://micromanager.win/download) [macOS](#macos) [Windows](#windows) [Linux](#linux) The Download button picks the build for the system you are browsing from; the sections below have every file. The newest version and its files are listed in each system’s release manifest, linked below. ## Is a version out yet? If a download button brought you back to this page, no release has been published for that system yet, or the newest one is still uploading. Leave your email below to hear when it is out. Each system’s manifest always names its current version: [latest-mac.yml](https://micromanager.win/releases/mac/latest-mac.yml), [latest.yml](https://micromanager.win/releases/win/latest.yml) for Windows and [latest-linux.yml](https://micromanager.win/releases/linux/latest-linux.yml) for Linux. ## How do I install microManager on a Mac? [Apple Silicon (DMG)](https://micromanager.win/download/mac)[Intel (DMG)](https://micromanager.win/download/mac-intel) Open the DMG and drag microManager into Applications. Or paste this into Terminal: `curl -fsSL https://micromanager.win/install.sh | sh` The install script: 1. checks it is on macOS and picks the Apple Silicon or Intel build, 2. reads the release manifest, `latest-mac.yml`, and downloads the matching zip, 3. checks the zip’s SHA-512 against the manifest and stops if they differ, 4. checks the app’s code signature and its Apple developer team, 5. copies the new app next to any existing `/Applications/microManager.app`, swaps them, and only then removes the old copy, asking first if microManager is running, and 6. opens microManager. It asks for your password (through `sudo`) only if your user cannot write to the install folder. [Read the script](https://micromanager.win/install.sh) before running it if you like; everything runs inside one function called on its last line, so a partly downloaded copy does nothing. Put a setting before `sh`, for example `curl -fsSL https://micromanager.win/install.sh | MICROMANAGER_NO_OPEN=1 sh`. The same settings work on Linux. | Setting | Effect | | ----------------------------------------- | -------------------------------------------------------------------------- | | `MICROMANAGER_INSTALL_DIR=~/Applications` | Install somewhere other than /Applications (Mac) or \~/.local/bin (Linux). | | `MICROMANAGER_NO_OPEN=1` | Do not open microManager when the install finishes. | | `MICROMANAGER_ASSUME_YES=1` | Mac: quit and replace a running copy without asking. | ### Is there a build for Intel Macs? Apple Silicon comes first. An Intel build is published only for releases where one is available; when there is none, the Intel button brings you back here. The install script picks the right build for your Mac by itself, and on an Apple Silicon Mac running a Terminal under Rosetta it still installs the Apple Silicon build. ### What does a Mac need? - macOS 13 Ventura or later, on a Mac with Apple Silicon or an Intel processor. An Intel build is published only for releases where one is available. - git, which comes with Apple’s Command Line Tools. - At least one coding agent CLI, installed and signed in: Claude Code, Codex, Grok Build, OpenCode, Gemini CLI or another ACP agent. - The GitHub CLI (gh), signed in, for pull request features. Everything else works without it. ## How do I install microManager on Windows? [Windows x64 (installer)](https://micromanager.win/download/windows) Paste this into PowerShell: `irm https://micromanager.win/install.ps1 | iex` It reads the release manifest, `latest.yml`, downloads the installer it names, checks the installer’s SHA-512 against the manifest and stops if they differ, then runs it for your user and deletes the download. [Read the script](https://micromanager.win/install.ps1) first if you like; everything runs inside one function called on its last line, and an error stops it without closing your PowerShell window. Set a setting first, for example `$env:MICROMANAGER_SILENT = '1'`, then paste the command. | Setting | Effect | | -------------------------------- | -------------------------------------------------------------------------------------------- | | `$env:MICROMANAGER_SILENT = '1'` | Install without showing the installer’s progress window. | | `$env:MICROMANAGER_NO_RUN = '1'` | Only download and check the installer, into the current folder, and leave running it to you. | Or install it by hand: 1. Download the installer and run it. It installs for your user only, so it does not ask for administrator rights. 2. The Windows installer is not code-signed yet, so Microsoft Defender SmartScreen may warn that it is from an unknown publisher. Check the file’s SHA-512 (below) first, then choose **More info** and **Run anyway**. 3. Open microManager from the Start menu. Setup, run and archive scripts run with Git Bash when Git for Windows is installed, so the same bash scripts work, otherwise PowerShell. Claude Code uses Git Bash for its own shell commands when it is there, too. Shortcuts use `Ctrl` where the Mac uses `⌘`; see [keyboard shortcuts](https://micromanager.win/docs/keyboard-shortcuts). ### What does a Windows PC need? - A 64-bit x64 PC running Windows. Windows on ARM is not supported yet. - Git for Windows, which gives microManager git and the Git Bash your scripts run in. Claude Code runs its own shell commands in Git Bash too when it is there, and in PowerShell when it is not. - At least one coding agent CLI, installed and signed in: Claude Code, Codex, Grok Build, OpenCode, Gemini CLI or another ACP agent. - The GitHub CLI (gh), signed in, for pull request features. Everything else works without it. ## How do I install microManager on Linux? [Linux x64 (AppImage)](https://micromanager.win/download/linux)[Debian and Ubuntu x64 (.deb)](https://micromanager.win/download/linux-deb) Paste this into a terminal to install the AppImage: `curl -fsSL https://micromanager.win/install.sh | sh` The install script: 1. reads the release manifest, `latest-linux.yml`, and downloads the AppImage, 2. checks its SHA-512 against the manifest and stops if they differ, 3. installs it as `~/.local/bin/micromanager`, replacing any previous copy in one step, 4. adds microManager to your applications menu with its icon, under `~/.local/share`, and 5. opens it. It never uses `sudo`, and the installed AppImage updates itself. To do it by hand instead, make the AppImage executable and start it; it runs on most distributions without installing anything: ```sh chmod +x microManager-*.AppImage ./microManager-*.AppImage ``` On Debian, Ubuntu and their derivatives, install the .deb with your package manager instead: ```sh # Use the name of the file you downloaded; apt needs the ./ to treat it as a file sudo apt install ./FILE.deb ``` Setup, run and archive scripts run with your login shell when it is zsh or bash, otherwise bash, in the workspace folder, with the same [environment variables](https://micromanager.win/docs/environment-variables) as on every other system. ### What does a Linux computer need? - A 64-bit x64 Linux desktop with glibc 2.31 or newer: Ubuntu 20.04, Debian 11, Fedora 32 or later. ARM64 Linux is not supported yet. - git, from your distribution’s packages. - At least one coding agent CLI, installed and signed in: Claude Code, Codex, Grok Build, OpenCode, Gemini CLI or another ACP agent. - The GitHub CLI (gh), signed in, for pull request features. Everything else works without it. ## How can I check a download myself? Every release manifest lists each file’s SHA-512 in base64. Compute the same value for the file you downloaded and compare. ### macOS ```sh # The zip's SHA-512, in the base64 form latest-mac.yml uses openssl dgst -sha512 -binary microManager-*-arm64.zip | openssl base64 -A # Who signed the app, and whether Gatekeeper accepts it codesign -dv --verbose=2 /Applications/microManager.app 2>&1 | grep TeamIdentifier spctl -a -vv /Applications/microManager.app ``` Compare the first result with the `sha512` next to the zip in [latest-mac.yml](https://micromanager.win/releases/mac/latest-mac.yml). Mac releases are signed with a Developer ID and notarized by Apple. ### Windows ```powershell # PowerShell: the installer's SHA-512 in the base64 form latest.yml uses $file = Get-Item .\microManager-Setup-*.exe $sha = [System.Security.Cryptography.SHA512]::Create() [Convert]::ToBase64String($sha.ComputeHash([IO.File]::ReadAllBytes($file.FullName))) ``` Compare it with the `sha512` in [latest.yml](https://micromanager.win/releases/win/latest.yml). ### Linux ```sh # The AppImage's SHA-512 in the base64 form latest-linux.yml uses openssl dgst -sha512 -binary microManager-*.AppImage | openssl base64 -A ``` Compare it with the `sha512` in [latest-linux.yml](https://micromanager.win/releases/linux/latest-linux.yml). ## How does updating work? The Mac app, the Windows app, the Linux AppImage and the .deb check for a new version when they start and every six hours, download it in the background and show **Restart to update** when it is ready. Updating the .deb asks for your password, because it installs the new package for the whole system. Running the install command again also installs the newest version; on Linux it installs the AppImage, so if you use the .deb, let it update itself instead. ## How do I uninstall it? Quit microManager first. Its own data (workspace records, chats and settings) lives in `~/Library/Application Support/microManager` on macOS , `%APPDATA%\microManager` on Windows and `~/.config/microManager` on Linux ; delete that folder too if you want a clean slate. - macOS: delete the app and, if you want, its data: ```sh rm -rf /Applications/microManager.app rm -rf ~/Library/Application\ Support/microManager ``` - Windows: Settings → Apps → Installed apps (Apps & features on Windows 10) → microManager → Uninstall. - Linux: delete the AppImage (the install command puts it at `~/.local/bin/micromanager`, with `~/.local/share/applications/micromanager.desktop`), or remove the .deb with `sudo apt remove micromanager`. Workspaces themselves are git worktrees, by default under `~/micromanager/workspaces` in your home folder. Commit or push anything you need from them before deleting that folder, then run `git worktree prune` in each repository. --- title: "Getting started with microManager" description: "Install microManager on macOS, Windows or Linux, import your Conductor setup or add a repository, and start your first agent in its own git worktree." url: "https://micromanager.win/docs/getting-started" updated: "2026-11-09" --- # Getting started Install microManager on macOS, Windows or Linux, then import your Conductor setup in one click (on a Mac) or add a git repository. Each new workspace gets its own git worktree and branch, and the agent you pick starts working there. Updated 9 November 2026 ![The first-run screen in the light theme: Welcome to microManager, buttons for Import from Conductor and Add a repository, and the agents it found, 3 of 5 installed.](https://micromanager.win/screenshots/onboarding-first-run.png) The first run offers the Conductor import and lists the agents it found. ## What do I need before installing? - A Mac with macOS 13 Ventura or later (Apple Silicon, or Intel for releases with an Intel build), a Windows 10 or 11 PC with an x64 processor, or an x64 Linux computer. - `git`. On a Mac, if `git --version` asks you to install the Command Line Tools, accept. On Windows, install Git for Windows, whose Git Bash runs your scripts and, when it is there, Claude Code’s shell commands. On Linux, install `git` from your distribution’s packages. - At least one coding agent installed and signed in: [Claude Code](https://micromanager.win/agents/claude-code), [Codex](https://micromanager.win/agents/codex), [Grok Build](https://micromanager.win/agents/grok), [OpenCode](https://micromanager.win/agents/opencode), [Gemini CLI](https://micromanager.win/agents/gemini-cli) or another ACP agent. microManager runs the CLIs you already have; it does not bundle them or sign you in to them. - The GitHub CLI (`gh`), signed in with `gh auth login`, if you want to create, check and merge pull requests from the app. There is no microManager account to create. ## How do I install microManager? The [download page](https://micromanager.win/install) has every file, the requirements for each system and how to check a download yourself. ### On a Mac Download the [Apple Silicon DMG](https://micromanager.win/download/mac) (or the [Intel DMG](https://micromanager.win/download/mac-intel) when a release has one), open it and drag microManager into Applications. Mac releases are signed with a Developer ID and notarized by Apple. Or install from Terminal: ```sh curl -fsSL https://micromanager.win/install.sh | sh ``` The script, which you can [read before you run it](https://micromanager.win/install.sh): 1. checks that it is running on macOS and picks the Apple Silicon or Intel build, 2. downloads the newest zip from `micromanager.win/releases/mac/`, 3. checks the zip’s SHA-512 against the release manifest `latest-mac.yml`, 4. checks the app’s code signature and its Apple developer team, 5. installs it as `/Applications/microManager.app`, asking first if a copy is running, and 6. opens it. It uses `sudo` only if your user cannot write to `/Applications`. Set `MICROMANAGER_INSTALL_DIR` to install somewhere else, such as `~/Applications`, and `MICROMANAGER_NO_OPEN=1` to skip opening the app at the end. ### On Windows Paste this into PowerShell: ```powershell irm https://micromanager.win/install.ps1 | iex ``` The [script](https://micromanager.win/install.ps1) downloads the newest installer from `micromanager.win/releases/win/`, checks its SHA-512 against the release manifest `latest.yml`, and runs it for your user only, without administrator rights. Or download the [Windows installer](https://micromanager.win/download/windows) and run it yourself. The installer is not code-signed yet, so SmartScreen may warn about an unknown publisher; the [download page](https://micromanager.win/install#windows) shows how to check the file first. ### On Linux Paste this into a terminal: ```sh curl -fsSL https://micromanager.win/install.sh | sh ``` The same [script](https://micromanager.win/install.sh) as on a Mac downloads the newest AppImage from `micromanager.win/releases/linux/`, checks its SHA-512 against `latest-linux.yml`, installs it as `~/.local/bin/micromanager` and adds it to your applications menu, all without `sudo`. Or download the [AppImage](https://micromanager.win/download/linux), make it executable with `chmod +x` and run it, or install the [.deb](https://micromanager.win/download/linux-deb) on Debian or Ubuntu with `sudo apt install ./FILE.deb`. ## What happens the first time I open it? With no repositories added yet, microManager shows which agent CLIs it found on your `PATH` (your login shell’s on macOS and Linux) and whether each one runs. On a Mac where it finds Conductor’s data, it offers **Import from Conductor**; otherwise, or as well, choose **Add a repository**. ## How do I bring over my Conductor setup? On the Mac where you used Conductor, choose **Import from Conductor**. microManager reads Conductor’s local database read-only and shows a preview: your repositories, each with a switch to include it, how many workspaces and chats it found, and whether it found your Conductor settings. Import keeps every workspace where it is on disk, so your branches and uncommitted work stay exactly as they were. The [migration guide](https://micromanager.win/migrate-from-conductor) covers each step and what carries over. ## How do I add a repository? Choose **Add a repository** (or press **⌘O**, Ctrl+O on Windows and Linux), then either pick a folder on your computer that is a git repository, or choose **Clone from URL**, paste the URL and pick the folder to clone it into. microManager reads the repository’s default branch and any settings files it contains, including Conductor’s `.conductor/settings.toml` ([repository settings](https://micromanager.win/docs/repository-settings)). ## How do I start my first workspace? Press **⌘N** (Ctrl+N on Windows and Linux), or **New workspace** in the sidebar, and choose: - the repository, - the branch to start from (searchable; the repository’s default branch unless you pick another), - optionally, a name (otherwise the workspace is named from your first message), - the agent, its model, effort and permission mode, and - optionally, a first message for the agent. microManager then creates a git worktree on a new branch, copies your ignored files such as `.env` into it, reserves ten ports for it, runs your setup script in the background and opens a chat with the agent. The [workspaces page](https://micromanager.win/docs/workspaces) explains each of those steps. ## How do I work with the agent? Type in the message box at the bottom of the chat. You can paste or drag in images, mention files with `@` (with fuzzy search over the worktree), and use the agent’s own slash commands, which pass straight through. Pickers along the bottom of the box set the model, the effort level and the permission mode (**Ask first**, **Accept edits**, **Plan**, **Auto** or **Bypass permissions**). A message you send while Claude Code or Codex is working is added to the running turn, and the agent reads it at its next step; **Alt+Enter** queues it for when the turn ends instead (other agents always get it queued), and **Send now** on a queued message interrupts the turn. Press **Esc** in an empty message box to stop the running turn. When the agent asks permission to run a tool, or asks you a question, a card appears in the chat for your answer. Open more tabs with **⌘T** (Ctrl+T) to run several sessions, even different agents, in the same workspace. ## How do I review and ship the work? The **Changes** panel lists every changed file with its added and removed lines, against the branch it will merge into (its target branch) or just the uncommitted work. Open a file to read the diff, click a line number to leave a comment, then **Send to agent** (it counts your open comments: **Send 2 to agent**). When you are happy, **Create PR** asks the agent to commit, push and open the pull request with `gh`; the **Checks** panel follows its CI, and **Merge…** in the pull request button’s menu merges it. See [diff review](https://micromanager.win/features/diff-review) and [pull requests](https://micromanager.win/features/pull-requests). ## How does microManager update itself? It checks `micromanager.win/releases/` for its system when it starts and every six hours, downloads a new version in the background, and shows **Restart to update** when it is ready. That covers the Mac app, the Windows app, the Linux AppImage and the .deb; updating the .deb asks for your password, because it installs the new package for the whole system. ## Where to go next - [Workspaces, branches and archiving](https://micromanager.win/docs/workspaces) - [Move from Conductor step by step](https://micromanager.win/migrate-from-conductor) - [Agents and profiles](https://micromanager.win/docs/agents) --- title: "Workspaces: worktrees, branches, ports and archiving" description: "How microManager creates a workspace (worktree, branch, files, ports, setup script), names it, tracks its status, archives it and restores it." url: "https://micromanager.win/docs/workspaces" updated: "2026-11-08" --- # Workspaces A workspace is one task in one repository. It is a git worktree on its own branch, with its own chat tabs, terminals, ten reserved ports and a setup script run, and it can be archived without losing its branch, its uncommitted work or its .context notes. Updated 8 November 2026 ![The New workspace dialog over the app: repository ledger-api, a new branch from main, a name to be set from the first message, agent Claude Code, model Haiku, effort Agent default, mode Accept edits, and a first message asking it to reject account names longer than 60 characters.](https://micromanager.win/screenshots/new-workspace-dialog.png) A new workspace is a git worktree on its own branch, with an agent ready to work in it. ## What happens when I create a workspace? When you press **⌘N** (Ctrl+N on Windows and Linux) and pick a repository, a base branch and an agent, microManager: 1. **Picks a name.** It chooses a city from a bundled list of about 400 world cities that no other workspace or branch in the repository uses, such as `lisbon`. The branch is `/lisbon`, where the prefix comes from Settings → General → Branch prefix: your GitHub username (looked up once with `gh api user`), a custom prefix, or none. 2. **Creates the worktree.** It fetches the base branch from the remote (giving up after 30 seconds, so it still works offline) and runs `git worktree add -b /`. If the remote has no such branch, it starts from your local one. The new branch has no upstream until it is pushed. 3. **Adds a `.context/` folder** to the worktree and lists it in the repository’s git exclude file (`.git/info/exclude`), so notes and screenshots shared between agents are never committed. 4. **Copies your ignored files**, such as `.env`, from the repository’s main checkout. See [files to copy](https://micromanager.win/docs/files-to-copy). 5. **Reserves ten ports**: the lowest multiple of 10 from 51000 up that no other active workspace uses. The first one is in `MICROMANAGER_PORT` (and `CONDUCTOR_PORT`). See [environment variables](https://micromanager.win/docs/environment-variables). 6. **Runs your setup script** in a **Setup** terminal you can watch. The workspace is ready as soon as the worktree exists, so agents can start while setup is still running. While it runs, and if it fails, a bar above the chat says so, with **Show output**; its log is kept, so you can read it after a restart. 7. **Opens the first chat** with the agent and model you picked, and sends your first message if you wrote one. To work on a branch that already exists, or on an open pull request, choose **Branch** or **Pull request** under **Start from** in the same dialog, or press **⌘⇧N** (Ctrl+Shift+N). The worktree then checks that branch out instead of making a new one: a branch only on the remote is checked out tracking it, and a pull request’s branch comes through `gh pr checkout`, with the workspace named after the pull request and targeting the branch it merges into. ## Where do workspaces live on disk? By default in `~/micromanager/workspaces//`, inside your home folder on every system. Change the folder in Settings → General → Workspaces folder; new workspaces go to the new place. Workspaces imported from Conductor stay exactly where they already are, and nothing is moved. ## How does automatic naming work? After you send the first message in a workspace’s first chat, microManager can give the workspace a readable name and its branch a matching one, such as “Add CSV export” on `/add-csv-export`. It does this only when: - **Name workspaces from the first message** is on in Settings → General (it is by default), - the branch still has its placeholder city name, and - the branch has no upstream, is not on the remote and has no open pull request. A name you typed yourself, in the New workspace dialog or later, is kept; only the branch is renamed then. It asks Claude Code for a name in one-shot mode with the Haiku model (giving up after 20 seconds) when Claude Code is available, whichever agent the chat runs, and otherwise derives one from your message on your computer. The branch is renamed with `git branch -m`. A branch that has been pushed or has an open pull request is never renamed. You can rename the workspace from the workspace header at any time, and its branch there too until it is pushed. ## What do the sidebar’s marks mean? Each workspace in the sidebar shows: - **Activity**: a spinner while the workspace is being set up or an agent is working, a mark when an agent needs your answer, an error mark, a dot for unread replies, or a check once the workspace is Done or its pull request is merged. - **Branch**, **diff size** (lines added and removed) and a **pull request badge**: its number, colored by its checks while it is open, or Draft, Merged or Closed. Right-click a workspace to pin it, rename it, mark it read, copy its branch name, archive it or set your own status: Backlog, In progress, In review or Done. Pinned workspaces stay at the top; the rest are sorted by most recent activity. **⌘1** to **⌘9** (Ctrl+1 to Ctrl+9 on Windows and Linux) jump to the first nine. ## What is the target branch? The branch the workspace will merge into, shown as **→ target** in the header. Diffs compare against it and pull requests are opened against it. It starts as the repository’s default branch; pick another from the header at any time. ## What does archiving do? Archiving a finished workspace: 1. stops the workspace’s agents and terminals, 2. runs the archive script, if there is one (up to 2 minutes; one from the repository’s own files only once you trust it), 3. saves the uncommitted work, untracked and staged files included, as a hidden git commit under `refs/micromanager/archived/`, 4. copies the `.context/` folder into microManager’s own data folder, 5. records the commit the branch points to, 6. removes the worktree folder with `git worktree remove --force`, and 7. deletes the workspace’s checkpoints. **The branch is kept, and so are your uncommitted work and the notes in `.context/`.** Other files git ignores, such as `node_modules` or `.env`, are removed with the folder; restoring copies your files to copy again. With **Archive after merging** on in Settings → General, a workspace is archived as soon as you merge its pull request from microManager. ## How do I restore or delete an archived workspace? **Restore** recreates the worktree for the same branch at the same path (from the recorded commit if the branch was deleted meanwhile), puts the saved uncommitted work and the `.context/` folder back, copies your ignored files again and runs the setup script again. If the branch gained commits that change the same lines while the workspace was archived, the saved changes are kept aside instead of being forced over them, and a bar under the workspace’s tabs lists their files: **Bring back** puts them in the worktree, with conflict markers where both changed the same lines, and **Discard…** deletes them. **Delete** is available only for archived workspaces: it deletes the branch if it is merged into the target branch, or its pull request was merged and every local commit was pushed (a squash merge leaves no other trace), or when you confirm that you want to force it, and drops the saved work. ## What is the `.context` folder for? `.context/` sits inside every workspace and is ignored by git. Agents are told about it, so it is a good place for plans, notes and screenshots that several agents, or several tabs, should share without committing them. ## Where to go next - [Repository settings](https://micromanager.win/docs/repository-settings) - [Files to copy into new workspaces](https://micromanager.win/docs/files-to-copy) - [Why worktrees (feature overview)](https://micromanager.win/features/git-worktrees) --- title: "Repository settings: settings.toml keys | microManager" description: "Every repository settings key microManager reads, which settings file wins, and how Conductor's settings.toml and conductor.json are read." url: "https://micromanager.win/docs/repository-settings" updated: "2026-11-09" --- # Repository settings microManager reads a repository’s scripts, files-to-copy list, environment variables and prompts from TOML files in the repository, using the same keys as Conductor. Its own .micromanager files take precedence over .conductor ones, local files over shared ones, and overrides made in the app over everything. Updated 9 November 2026 ## Which files does microManager read, and in what order? It merges up to seven sources. When two set the same key, the one higher in this list wins, and keys a source does not mention fall through to the next one. Run scripts merge by their id across the files, while run scripts set in the app replace the files’ list. | Precedence | Source | Notes | | ---------- | ---------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | 1 | Overrides made in microManager’s repository settings | Stored in microManager, never written to your repository. | | 2 | `.micromanager/settings.local.toml` | Personal overrides; keep it out of git. | | 3 | `.micromanager/settings.toml` | Shared, committed settings for microManager. | | 4 | `.conductor/settings.local.toml` | Conductor’s personal overrides file, read as is. | | 5 | `.conductor/settings.toml` | Conductor’s shared settings file, read as is. | | 6 | `conductor.json` | Conductor’s older JSON format: scripts.setup, scripts.run (a string), scripts.archive and runScriptMode. Ignored when .conductor/settings.toml exists, as in Conductor. | | 7 | Built-in defaults | No scripts; files to copy default to .env\*. | microManager never writes to any of these files. Changes you make in a repository’s settings in the app are stored in microManager’s own database, on your computer. ## Which keys can a settings file contain? | Key | Type | What it does | | ------------------------------------ | ------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `scripts.setup` | string | Shell command run in a new workspace right after its worktree exists (and again when you restore an archived workspace). Agents can start while it runs. | | `scripts.archive` | string | Shell command run before a workspace is archived, while its folder still exists. It gets up to 2 minutes, and a failure does not stop the archive. | | `scripts.run_mode` | "concurrent" \| "nonconcurrent" | With "nonconcurrent", starting a run script stops the same script in the repository’s other workspaces first. Useful when every workspace wants the same fixed port. | | `scripts.run..command` | string | The command of one run script. The Run button in the workspace header starts the default script, and its menu lists every script. | | `scripts.run..default` | boolean | Makes this the script the Run button starts. With none marked, the first script is the default. | | `scripts.run..args` | string\[] | Arguments added after the command, each quoted for the shell the script runs in. | | `scripts.run..options.cwd` | string | The folder the script runs in, relative to the workspace. The workspace folder when absent. | | `scripts.run..icon` | string | Accepted for compatibility with Conductor’s format; microManager does not show it. | | `scripts.run..available_in` | string \| string\[] | Where the script may run. A script limited to "cloud" is left out, because every microManager workspace is local. | | `scripts.run..hide` | boolean | Leaves the script out. Run scripts merge by id across the files, field by field, so a local file can hide a shared script with just this line. | | `scripts.run` | string (legacy) | A single command string is still accepted and becomes one script named default, marked as the default. | | `file_include_globs` | string\[], or a string with one pattern per line | Patterns, written like .gitignore lines, for untracked or ignored files to copy into new workspaces. Used when the repository has no .worktreeinclude file. | | `environment_variables.` | string | A variable for the scripts, terminals and agents of every workspace of the repository. Variables merge by name across the files, and never replace the workspace’s own MICROMANAGER\_\* and CONDUCTOR\_\* variables. | | `environment_variables.local.` | string | Wins over the same file’s environment_variables for local workspaces, which every microManager workspace is. The .cloud table is left out. | | `prompts.general` | string | Instructions added to every chat in the repository’s workspaces, after your own general prompt, once you trust the repository. | | `prompts.create_pr` | string | Replaces the prompt the Create PR button sends to the agent in this repository, once you trust it. The same goes for each prompt below. | | `prompts.code_review` | string | Replaces the prompt Ask for a review sends to a new Review chat. | | `prompts.fix_errors` | string | Replaces the prompt that asks the agent to fix the checks failing on the pull request. | | `prompts.resolve_merge_conflicts` | string | Replaces the prompt that asks the agent to merge the target branch and resolve conflicts. | | `prompts.rename_branch` | string | Replaces the prompt that asks the agent to give the branch a descriptive name. | | `run_from_root` | boolean | Run from repository root, for a project that only runs from the repository’s own folder: a workspace’s Run from root button then shows its changes there, kept in sync, so a dev server running there picks them up, and turning it off puts the folder back as it was. Off by default. Conductor’s key for the same setting, spotlight_testing, is read too. | ## What does a complete settings file look like? ```toml # .micromanager/settings.toml (or .conductor/settings.toml) file_include_globs = [".env", ".env.local", "config/secrets.json"] [scripts] setup = "pnpm install --frozen-lockfile" archive = "docker compose -p \"$MICROMANAGER_WORKSPACE_NAME\" down -v" run_mode = "concurrent" [scripts.run.dev] command = "pnpm dev --port $MICROMANAGER_PORT" default = true [scripts.run.tests] command = "pnpm test --watch" ``` The same file works as `.conductor/settings.toml`, so a repository that already has one needs no changes. ## How do I keep personal settings out of git? Put them in `settings.local.toml` next to the shared file and add it to `.gitignore`. It takes precedence over the shared file, so you can change one run script without touching what your team uses: ```toml # .micromanager/settings.local.toml (personal, keep it out of git) [scripts.run.dev] command = "pnpm dev --port $MICROMANAGER_PORT --inspect" default = true ``` ## Is the older conductor.json still read? Yes, at the lowest precedence of the files, unless the repository has `.conductor/settings.toml`, as in Conductor. It understands `scripts.setup`, a single `scripts.run` string, `scripts.archive` and `runScriptMode`: ```json { "scripts": { "setup": "npm install", "run": "npm run dev", "archive": "npm run clean" }, "runScriptMode": "nonconcurrent" } ``` ## Where do I see which file a setting came from? Open the repository in Settings → Repositories. Under Scripts, **Where these come from** lists the files in effect, highest precedence first. The setup, run and archive scripts, the run mode, the files to copy and Run from repository root can each be overridden there for this computer only, and the repository can have its own prompts. ## Where to go next - [Scripts in detail](https://micromanager.win/docs/scripts) - [Files to copy](https://micromanager.win/docs/files-to-copy) - [Environment variables](https://micromanager.win/docs/environment-variables) --- title: "Setup, run and archive scripts | microManager docs" description: "How microManager runs a repository's setup, run and archive scripts, which shell and environment they get, and how Conductor's scripts keep working." url: "https://micromanager.win/docs/scripts" updated: "2026-11-09" --- # Setup, run and archive scripts A repository can define a setup script that prepares each new workspace, run scripts that become Run buttons, and an archive script that cleans up. microManager reads them from the same TOML keys Conductor uses, so scripts you already have keep working. Updated 9 November 2026 ![A web-dashboard workspace whose bar says Setup did not run: web-dashboard is not trusted yet, under the Trust web-dashboard? dialog its Review… button opens, which lists the setup and archive scripts, the VITE_LEDGER_API_URL variable and the CLAUDE.md agents would read, with Cancel and Trust and run setup buttons.](https://micromanager.win/screenshots/repository-trust.png) A repository’s own scripts, variables and agent files wait until you trust it, and Review… lists what trusting turns on. ## Where do scripts come from? From the repository’s settings files, highest precedence first: overrides you make in microManager, `.micromanager/settings.local.toml`, `.micromanager/settings.toml`, `.conductor/settings.local.toml`, `.conductor/settings.toml`, then the older `conductor.json`. Repository settings in microManager show each script together with the file it came from. The full list of keys is on the [repository settings](https://micromanager.win/docs/repository-settings) page. ## Why did a script from the repository not run? Anyone can commit a settings file, and its scripts and variables run on your computer, so microManager uses what a repository’s own files ask for only once you trust that repository. Until then, new workspaces start without its setup script, archiving skips its archive script, its `environment_variables` are left out, Claude Code reads only your own settings, not the repository’s CLAUDE.md, `.claude` folder (hooks and permission rules) or `.mcp.json`, OpenCode reads none of its project config (`opencode.json`, the `.opencode` folder with its plugins) or instructions, and git runs none of the programs the repository’s own `.git/config` names (an fsmonitor hook, filter commands) or its hooks. A clone never brings those along, but a folder copied from elsewhere can. The **New workspace** dialog lists what the files ask for, with a switch to trust the repository, and a bar in each of its workspaces offers **Review…**, which shows the same list before you trust it. You can trust a repository, or stop trusting it, under Settings → Repositories → **Trust**. Codex, Gemini CLI and Grok Build keep their own list of folders you trust and ask about each one themselves. Scripts you set in microManager itself always run, run scripts start only when you press **Run**, and repositories imported from Conductor come in trusted. ## How does the setup script run? Right after a workspace’s worktree exists, and again when you restore an archived workspace, microManager runs `scripts.setup` in a **Setup** terminal you can open from the terminal drawer (**⌘J**, or Ctrl+J on Windows and Linux). It runs in the background: the workspace is usable at once and agents can start working while it runs. While it runs, and if it fails, a bar above the chat says so, with **Show output** and, after a failure, **Run again**. Its log is kept, so the drawer can show it after a restart, and **Run setup again** in the drawer (or **Run the setup script again** in the ⌘K command palette) runs it once more. ```toml # .conductor/settings.toml or .micromanager/settings.toml [scripts] setup = "pnpm install --frozen-lockfile" ``` ## How do run scripts work? Every `scripts.run.` table is a run script. The **Run** button in the workspace header starts the default one (marked `default = true`, otherwise the first) and becomes **Stop** while it runs; its menu starts or stops each script and has **Stop all run scripts**. ```toml [scripts.run.dev] command = "pnpm dev --port $MICROMANAGER_PORT" default = true [scripts.run.storybook] command = "pnpm storybook --port $((MICROMANAGER_PORT + 1))" ``` A run script runs in its own terminal tab, such as **Run dev**. On macOS and Linux, **Stop** sends SIGINT to the script’s whole process group and, if it is still running after 5 seconds, SIGKILL; on Windows it ends the script’s whole process tree. Agents can read the output with the [`GetTerminalOutput`](https://micromanager.win/docs/mcp-tools) tool, which helps when you ask one to fix a failing dev server. Run scripts merge by id across the settings files, field by field, so `.conductor/settings.local.toml` can change just one field of a shared script: `hide = true` leaves a script out of the Run menu on your machine, and `default = true` makes another one the default. ```toml # .conductor/settings.local.toml [scripts.run.storybook] hide = true ``` An older single-string form still works and becomes one script called `default`: ```toml [scripts] run = "npm run dev" ``` ## What does `run_mode` do? `scripts.run_mode = "nonconcurrent"` lets only one workspace of the repository run a given script at a time: starting it in one workspace first stops it in the others. Use it when the script needs a fixed port or another resource that cannot be shared. The default, `"concurrent"`, lets every workspace run it at once, which works well when each script uses its own `$MICROMANAGER_PORT`. ## What does the archive script do? `scripts.archive` runs when you archive a workspace, after its agents and terminals stop and before its folder is removed, with up to 2 minutes to finish and no terminal. If it fails, the archive still goes ahead. Use it to stop containers, drop a per-workspace database or delete caches outside the worktree. ```toml [scripts] archive = "docker compose -p \"$MICROMANAGER_WORKSPACE_NAME\" down -v" ``` ## Which shell and environment do scripts get? Scripts run in the workspace folder. On macOS and Linux they run with your `$SHELL -c` when it is zsh or bash, otherwise with `zsh -c` on macOS and `bash -c` on Linux, and they get the environment of your login shell, so tools installed with Homebrew, nvm, mise or asdf are on the `PATH`. On Windows they run in Git Bash when Git for Windows is installed, so the same bash scripts work, and otherwise in PowerShell, with the environment Windows gives apps. Every script also gets the workspace variables: `MICROMANAGER_WORKSPACE_NAME`, `MICROMANAGER_WORKSPACE_PATH`, `MICROMANAGER_ROOT_PATH`, `MICROMANAGER_DEFAULT_BRANCH`, `MICROMANAGER_PORT` and the rest, listed on the [environment variables](https://micromanager.win/docs/environment-variables) page. Agents and terminals get the same environment. Variables of your own go in an `[environment_variables]` table of any settings file. A file’s `[environment_variables.local]` table wins over its shared table, `[environment_variables.cloud]` is left out (every microManager workspace is local), and a higher file wins over a lower one, name by name. They come after your login environment and never replace the workspace variables. Settings → Repositories lists the ones in effect. ```toml [environment_variables] API_URL = "https://staging.example.com" [environment_variables.local] API_URL = "http://localhost:4000" ``` ## Will my Conductor scripts work unchanged? Yes, for local scripts. microManager reads `.conductor/settings.toml`, `.conductor/settings.local.toml` and `conductor.json` as they are, and with **Conductor-compatible environment** on (the default), it sets the `CONDUCTOR_*` variables Conductor gave local workspaces alongside their `MICROMANAGER_*` twins, with the meanings they had in Conductor: `$CONDUCTOR_WORKSPACE_NAME` is built from the workspace’s name the way Conductor built it, and `$CONDUCTOR_DEFAULT_BRANCH` is the target branch. Two things differ: an imported workspace gets a new block of ports, so `$CONDUCTOR_PORT` changes, and Conductor’s cloud and API variables (`CONDUCTOR_API_KEY`, `CONDUCTOR_SESSION_ID` and the like) are not set. Run actions with `args` and `options.cwd` work too, and an action limited to Conductor’s cloud is left out. microManager never writes to these files. On Windows and Linux, a script that calls a macOS-only command such as `open`, `pbcopy` or `brew` needs an equivalent there; put that version in `.micromanager/settings.local.toml` on the machine that needs it. ## Where to go next - [Every repository settings key](https://micromanager.win/docs/repository-settings) - [Environment variables scripts receive](https://micromanager.win/docs/environment-variables) --- title: "Environment variables (MICROMANAGER_*, CONDUCTOR_*)" description: "The environment variables microManager sets for scripts, terminals and agents, their Conductor-compatible CONDUCTOR_* names, and how ports are reserved." url: "https://micromanager.win/docs/environment-variables" updated: "2026-11-08" --- # Environment variables microManager gives every script, terminal and agent in a workspace the same set of variables: the workspace’s name, id and path, the repository’s path and default branch, and the first of ten reserved ports. Each also exists under its Conductor name, such as CONDUCTOR_PORT, with the meaning it had in Conductor, so existing scripts keep working. Updated 8 November 2026 ## Which variables does microManager set? | Variable | Conductor-compatible name | Value | Example | | ----------------------------- | -------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------- | | `MICROMANAGER` | None | Always 1, so a script can tell it is running inside microManager. | `1` | | `MICROMANAGER_WORKSPACE_NAME` | `CONDUCTOR_WORKSPACE_NAME` | The workspace’s folder name, a city such as lisbon. It stays the same when you rename the workspace or its branch. **CONDUCTOR_WORKSPACE_NAME:** The workspace’s name the way Conductor gave it to scripts: lowercase, spaces as hyphens, other punctuation left out. It changes when you rename the workspace, as in Conductor. | `lisbon`, `fix-login-bug` | | `MICROMANAGER_WORKSPACE_ID` | `CONDUCTOR_WORKSPACE_ID` | The workspace’s unique id. | `8f14e45f-ceea-467f-a0e7-4f3b7d9c1a2b` | | `MICROMANAGER_WORKSPACE_PATH` | `CONDUCTOR_WORKSPACE_PATH` | Absolute path of the workspace’s git worktree. | `~/micromanager/workspaces/my-app/lisbon` | | `MICROMANAGER_ROOT_PATH` | `CONDUCTOR_ROOT_PATH` | Absolute path of the repository’s main checkout, where files to copy come from. | `~/code/my-app` | | `MICROMANAGER_DEFAULT_BRANCH` | `CONDUCTOR_DEFAULT_BRANCH` | The repository’s default branch. **CONDUCTOR_DEFAULT_BRANCH:** The branch the workspace merges into (its target branch), as in Conductor. | `main`, `release/2.0` | | `MICROMANAGER_PORT` | `CONDUCTOR_PORT` | The first of ten ports reserved for this workspace (PORT to PORT+9). Each workspace gets the lowest free multiple of 10 from 51000 up. | `51000` | | `MICROMANAGER_IS_LOCAL` | `CONDUCTOR_IS_LOCAL` | Always 1: workspaces run on your own computer. | `1` | ## Who gets these variables? The setup, run and archive scripts, every terminal you open in the workspace, and every agent session. They are added to the environment of your login shell, so everything else on your `PATH` is there too. ## Can a repository add variables of its own? Yes, in an `[environment_variables]` table of its settings files, as in Conductor; a file’s `[environment_variables.local]` table, for local workspaces (every microManager workspace is one), wins over its shared table. They reach the same scripts, terminals and agents, and never replace the variables above. [Scripts](https://micromanager.win/docs/scripts) has an example. ## How do the Conductor-compatible names work? With **Conductor-compatible environment** on in Settings → General (it is on by default), each variable in the second column is set too, with the meaning it had in Conductor. Most hold the same value as their `MICROMANAGER_*` twin. Two differ: `CONDUCTOR_WORKSPACE_NAME` is built from the workspace’s name, as Conductor built it, so a Docker Compose project or database a Conductor script named after it keeps its name in an imported workspace; and `CONDUCTOR_DEFAULT_BRANCH` is the workspace’s target branch. Use `MICROMANAGER_WORKSPACE_NAME` for a name that never changes. Imported workspaces get a new block of ports, so `CONDUCTOR_PORT` differs from the one Conductor gave them. Conductor’s cloud and API variables, such as `CONDUCTOR_API_KEY` and `CONDUCTOR_SESSION_ID`, are not set: microManager has no cloud or Conductor API. Turn the setting off if you have finished moving your scripts to the new names. ## How are ports reserved? Each active workspace gets a block of ten ports. `MICROMANAGER_PORT` is the first, a multiple of 10 from 51000 up that no other active workspace holds, so `$MICROMANAGER_PORT` to `$((MICROMANAGER_PORT + 9))` are yours. Archiving a workspace frees its block for the next one. ```sh # Start each workspace's dev server on its own port pnpm dev --port "$MICROMANAGER_PORT" # A second service on the next reserved port pnpm storybook --port "$((MICROMANAGER_PORT + 1))" # Copy something from the main checkout that files-to-copy does not cover cp -R "$MICROMANAGER_ROOT_PATH/fixtures/large" ./fixtures/large ``` ## How can a script tell it is running in microManager? ```sh if [ "${MICROMANAGER:-}" = 1 ]; then echo "running in microManager workspace $MICROMANAGER_WORKSPACE_NAME" fi ``` ## Where to go next - [Scripts](https://micromanager.win/docs/scripts) - [Repository settings](https://micromanager.win/docs/repository-settings) - [Migrating from Conductor](https://micromanager.win/migrate-from-conductor) --- title: "Copy .env and other ignored files into worktrees" description: "Which ignored files microManager copies into new workspaces, how .worktreeinclude, file_include_globs and .env* are chosen, and what is never overwritten." url: "https://micromanager.win/docs/files-to-copy" updated: "2026-11-08" --- # Files to copy A new git worktree has none of your ignored files, so microManager copies them in from the repository’s main checkout. It uses .worktreeinclude when the repository has one, otherwise file_include_globs from the repository settings, otherwise every .env\* file. Updated 8 November 2026 ## Why do I need to copy files into a worktree? `git worktree add` checks out only what is committed. Files your `.gitignore` excludes, such as `.env`, local certificates or a `config/secrets.json`, are not there, so a dev server or test run in the new worktree fails until you copy them. microManager does this for you every time it creates or restores a workspace. ## How does microManager decide which files to copy? It uses the first of these that the repository has: | Order | Source | Format | | ----- | ----------------------------------------------- | ----------------------------------------------- | | 1 | `.worktreeinclude` in the repository root | One pattern per line, written like `.gitignore` | | 2 | `file_include_globs` in the repository settings | A list of glob patterns in TOML | | 3 | Built-in default | `.env*` | Only one source is used: a `.worktreeinclude` file replaces `file_include_globs` rather than adding to it. ## What does a `.worktreeinclude` file look like? ```text # .worktreeinclude .env .env.local config/secrets.json certs/*.pem ``` Lines follow `.gitignore` rules: `#` starts a comment, `*` matches within a folder and `**` across folders. The same file name is read by other tools, including Claude Code’s desktop app and Codex in the ChatGPT desktop app, so one file can serve all of them. ## How do I set `file_include_globs`? ```toml # .conductor/settings.toml or .micromanager/settings.toml file_include_globs = [".env", ".env.*", "apps/*/.env.local"] ``` Conductor’s key and format are read as they are, including its other form, one string with a pattern per line. A personal list can go in `settings.local.toml`, which takes precedence over the shared file. ## Which files are never copied or overwritten? - **Only ignored or untracked files are copied.** A file that git tracks already comes with the checkout. - **Nothing is overwritten.** If a file already exists in the new worktree, microManager leaves it alone. - **Ignored folders are searched only when a pattern points into them.** `.env*` does not pull `.env` files out of `node_modules`, while `config/**` does reach into an ignored `config/` folder. - **Nothing is copied through a linked folder.** If a folder in the new worktree is a symlink, files under it are skipped, so a copy never lands outside the worktree. - Files come from the repository’s **main checkout** (the folder you added, `$MICROMANAGER_ROOT_PATH`), never from another workspace. ## Can a setup script do the rest? Yes. For anything copying cannot do, such as generating a file or linking a shared folder, use the [setup script](https://micromanager.win/docs/scripts), which runs after the files are copied and can read `$MICROMANAGER_ROOT_PATH`. ## Where to go next - [Repository settings keys](https://micromanager.win/docs/repository-settings) - [Workspaces](https://micromanager.win/docs/workspaces) --- title: "Agents and profiles | microManager docs" description: "The five built-in agent profiles (Claude Code, Codex, Grok Build, OpenCode, Gemini CLI), what a profile holds, and how to add any ACP agent." url: "https://micromanager.win/docs/agents" updated: "2026-11-08" --- # Agents and profiles An agent profile tells microManager how to start one coding agent, with its kind, command, arguments, environment, models and effort levels. Five profiles are built in (Claude Code, Codex, Grok Build, OpenCode and Gemini CLI), and you can add any other agent that speaks the Agent Client Protocol. Updated 8 November 2026 ![Settings, Agents: the default agent set to Claude Code, an empty Favorites list with Add favorite, and the profiles Claude Code, Codex and Grok found with their versions and OpenCode and Gemini CLI not found, with Detect all and Add ACP agent buttons.](https://micromanager.win/screenshots/settings-agents.png) Settings → Agents shows which agent CLIs microManager found and their versions. ## Which agent profiles are built in? | Profile | Kind | Command | How microManager talks to it | | --------------------- | ------ | ------------------ | --------------------------------------------------------------------------- | | Claude Code | claude | `claude` | The Claude Agent SDK, pointed at your own `claude` CLI or wrapper script | | Codex | codex | `codex` | `codex app-server` over stdio (JSON-RPC); `codex exec --json` as a fallback | | Grok (for Grok Build) | acp | `grok agent stdio` | Agent Client Protocol, natively | | OpenCode | acp | `opencode acp` | Agent Client Protocol | | Gemini CLI | acp | `gemini --acp` | Agent Client Protocol | Built-in profiles can be edited and turned off, but not deleted. ## What does a profile contain? - **Kind**: `claude`, `codex` or `acp`, which decides how microManager talks to the agent. - **Command**: an absolute path, or a name looked up on your login shell’s `PATH`. - **Arguments** and **environment variables** added when the agent starts. - **Models**, one id per line, with a default model, and a **default effort** from the levels its kind offers. - **On or off**, so you can hide an agent you do not use. ## How does microManager find each agent’s command? It captures the environment of your login shell, and captures it again when you choose **Detect** or **Detect all**, so a `claude` or `codex` installed with Homebrew, npm, nvm or mise is found the way your terminal finds it. Settings → Agents shows, for each profile, **Found** with its version (hover for the resolved path), **Not found**, or **Not working** when the command exists but its version check fails (hover for the reason). **Detect** next to a profile’s command checks it again, and **Detect all** checks every profile. For Codex, when `codex` turns out to be the npm launcher script, microManager runs the native Codex binary that the package installs beside it. You can point a profile at a wrapper script instead of the agent itself, for example one that sets a proxy or picks an account. When you import from Conductor, the Claude Code, Codex and OpenCode paths set in Conductor’s settings become the commands of those profiles. ## Which models and effort levels can I pick? - **Claude Code**: `opus`, `sonnet`, `haiku` and `fable`, plus any model id you type. Effort: low, medium, high, xhigh and max. - **Codex**: the models in Codex’s own model cache (`~/.codex/models_cache.json`), refreshed from the app-server once a Codex chat starts, plus any model id you type; your `config.toml` model and effort are the defaults. Effort: low, medium, high, xhigh, max and ultra, as far as the chosen model supports them. - **ACP agents**: the models each agent offers over the protocol. The model and effort pickers sit along the bottom of the message box, so you can change them per chat. ## What are the permission modes? Each chat has a mode, picked at the bottom of the message box: **Ask first** (the agent asks before editing files or running commands), **Accept edits** (it edits files in the workspace without asking, and asks before commands and before writing anywhere else), **Plan** (it reads and plans, and changes nothing), **Auto** (a reviewer model approves or blocks each action for you: Claude Code’s auto mode and Codex’s auto review) and **Bypass permissions** (it runs everything without asking). New chats start in the agent’s own default: for Claude Code that is `permissions.defaultMode` from its settings (the workspace’s `.claude/settings.local.json`, then `.claude/settings.json`, then your own `~/.claude/settings.json`), otherwise Ask first. To pick one yourself, set **New chats start in** on the agent in Settings → Agents. When an agent asks, a card in the chat shows what it wants to do with the agent’s own choices: Claude Code offers **Allow**, **Deny** and, when it suggests one, an always-allow rule such as **Allow all edits this session**; Codex offers **Allow once**, **Allow for this session** and **Reject**. ACP agents decide for themselves whether to ask, and have no Auto mode. microManager answers their requests by the mode: it allows everything in Bypass permissions and edits to files in the workspace in Accept edits, each time for that one action only. When an agent offers only an “always allow”, which it would keep after you change the mode back, you get the question instead. An agent whose own settings approve everything, as Grok Build can be set to, never asks. Plan is offered only to an agent with a plan mode of its own; for an agent with no modes at all, such as Grok Build, the mode picker says so. ## How do I add another ACP agent? In Settings → Agents, choose **Add ACP agent**: give it a name, the command that starts it in ACP mode and any arguments, then add environment variables or models in its profile if it needs them. It then appears in the agent picker for new chats and workspaces like the built-in ones. See [ACP agents](https://micromanager.win/features/acp-agents) for what the protocol gives you. ## Are my agents’ own settings respected? Yes. microManager starts the agent CLIs you installed and signed in to, with their normal configuration. Claude Code loads your user settings, and the repository’s project and local settings (`CLAUDE.md`, permissions, hooks and MCP servers) once you [trust the repository](https://micromanager.win/docs/scripts#why-did-a-script-from-the-repository-not-run); repositories imported from Conductor come in trusted. Codex uses your Codex configuration. microManager adds its own `micromanager` MCP server and a short note about the workspace to each session; it does not replace your setup. ## Where to go next - [Claude Code in microManager](https://micromanager.win/agents/claude-code) - [Codex in microManager](https://micromanager.win/agents/codex) - [ACP agents](https://micromanager.win/features/acp-agents) --- title: "MCP tools agents get in microManager" description: "The six tools of microManager's micromanager MCP server, such as GetWorkspaceDiff, DiffComment and GetDiffComments, and how agents connect to it." url: "https://micromanager.win/docs/mcp-tools" updated: "2026-11-08" --- # MCP tools Every agent session gets a small MCP server named micromanager, with tools to ask you a question, read the workspace’s diff, leave and read review comments, read terminal output and look up the workspace. The tools Conductor had are here under the same names, so prompts and habits that mention them carry over; GetWorkspaceInfo is microManager’s own. Updated 8 November 2026 ## Which tools does the micromanager server offer? | Tool | Input | What it does | | ------------------- | ------------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `AskUserQuestion` | `{ questions: { question, header?, options: { label, description? }[], multiSelect? }[] }` | Shows a question card in the chat and returns your answer. Offered to non-Claude agents; Claude Code uses its own built-in tool of the same name, which microManager renders the same way. | | `GetWorkspaceDiff` | `{ path?: string, stat?: boolean }` | Returns the workspace’s unified diff against the merge base with its target branch, including uncommitted and untracked files. Pass a path for one file, or stat for a summary. | | `DiffComment` | `{ comments: { file, line, endLine?, body }[] }` | Adds review comments from the agent to the Review panel, anchored to lines of the diff. | | `GetDiffComments` | `{ file?: string }` | Returns the review comments on the diff that are not resolved yet, yours and agents’, each with its file, lines and the code it is on, so “address my comments” works without sending them. Pass a file for one file’s. Reading them changes nothing. | | `GetTerminalOutput` | `{ terminal?: "run" \| "setup" \| , lines?: number }` | Reads the last lines of output from the run script (the default), the setup script or a terminal in the workspace. | | `GetWorkspaceInfo` | `{}` | Returns the workspace’s name, repository, folder, main checkout, branch, target branch, reserved ports, setup status and pull request. | ## What does a DiffComment call look like? Comments appear in the workspace’s Review panel next to yours, marked as coming from the agent. You can resolve them, or send the open ones to the chat you are in. ```json { "comments": [ { "file": "src/api/users.ts", "line": 42, "endLine": 48, "body": "This retries forever when the token is revoked; cap it at three attempts." } ] } ``` ## How do agents connect to it? - **Claude Code** gets the server in-process, through the Claude Agent SDK. - **ACP agents** that support HTTP MCP servers reach it over Streamable HTTP on `127.0.0.1`, at a random port, with a bearer token made for that one session. - **Codex**, and ACP agents that only speak stdio MCP, get a small bridge that forwards to that same endpoint, with the token in its environment rather than on its command line. ## Can other programs or web pages call these tools? No. The endpoint listens only on `127.0.0.1`, every request needs that session’s 256-bit token, and requests carrying an `Origin` header, which browsers add to requests web pages make, are refused. It is the only network port microManager opens; the app’s own interface talks to its backend in-process, not over HTTP. ## Do my own MCP servers still work? Yes. The micromanager server is added alongside the ones you configured for each agent. Claude Code, for example, still loads the MCP servers from your user settings, and from the repository’s `.mcp.json` and project settings once you [trust the repository](https://micromanager.win/docs/scripts#why-did-a-script-from-the-repository-not-run). ## Where to go next - [Diff review](https://micromanager.win/features/diff-review) - [Agents and profiles](https://micromanager.win/docs/agents) --- title: "Keyboard shortcuts | microManager docs" description: "Every microManager keyboard shortcut on macOS, Windows and Linux: new workspace, new tab, switching workspaces, the command palette, terminal and stop." url: "https://micromanager.win/docs/keyboard-shortcuts" updated: "2026-11-09" --- # Keyboard shortcuts On a Mac, press ⌘K for the command palette, ⌘N for a new workspace, ⌘T for a new chat tab and ⌘1 to ⌘9 to jump between workspaces. On Windows and Linux the same shortcuts use Ctrl instead of ⌘, as the full list below shows. Updated 9 November 2026 ![The command palette over a workspace: a search box reading Jump to a workspace or run an action, the four workspaces with their repository and branch, then actions for the workspace on screen such as Open in Finder, Open in VS Code and Run server.](https://micromanager.win/screenshots/command-palette.png) ⌘K (Ctrl+K on Windows and Linux) jumps to any workspace or runs an action on the one you are in. ## What are all the shortcuts? | macOS | Windows and Linux | Action | | ----------------- | --------------------------- | ------------------------------------------------------------------------------------------ | | `⌘N` | `Ctrl+N` | New workspace | | `⌘⇧N` | `Ctrl+Shift+N` | New workspace on a branch that already exists or on a pull request | | `⌘O` | `Ctrl+O` | Add a repository | | `⌘T` | `Ctrl+T` | New chat tab in the current workspace | | `⌘W` | `Ctrl+W` | Close the current chat tab | | `⌘⇧T` | `Ctrl+Shift+T` | Reopen the chat tab you closed last | | `⌘1 … ⌘9` | `Ctrl+1 … Ctrl+9` | Jump to a workspace by its position in the sidebar | | `⌃Tab / ⌃⇧Tab` | `Ctrl+Tab / Ctrl+Shift+Tab` | Next or previous chat tab | | `⌘K` | `Ctrl+K` | Command palette: jump to a workspace or run an action | | `⌘B` | `Ctrl+B` | Show or hide the sidebar | | `⌘J` | `Ctrl+J` | Show or hide the terminal drawer | | `⌘⇧E` | `Ctrl+Shift+E` | Show or hide the Changes panel | | `⌘⇧D` | `Ctrl+Shift+D` | Open the diff viewer on the first changed file, or close it | | `⌘⇧P` | `Ctrl+Shift+P` | Create a pull request | | `⌘⌥R` | `Ctrl+Alt+R` | Ask an agent to review the workspace’s changes | | `⌘⇧A` | `Ctrl+Shift+A` | Go to the next workspace that needs you | | `⌘⇧O` | `Ctrl+Shift+O` | Open the workspace in your default app: the one you chose, else the first editor installed | | `⌘,` | `Ctrl+,` | Settings | | `⌘/` | `Ctrl+/` | List every shortcut | | `Esc` | `Esc` | Stop the running turn (when the message box is focused and empty) | | `Enter or ⌘Enter` | `Enter or Ctrl+Enter` | Send the message (Settings → General → Send messages with chooses which) | On a Mac, ⌘ is Command, ⌃ is Control and ⇧ is Shift. Menus show the keys for the system you are on. ## What can the command palette do? `⌘K` (`Ctrl+K` on Windows and Linux) opens a search box that jumps to any workspace by name and runs actions: new workspace, new tab, start or stop a run script, run the setup script again, create a pull request or open its checks, open the workspace in another app, pin it, set its status, copy its branch name, archive it, open a Settings section, switch the theme and show or hide the panels. ## How do I choose whether Enter sends a message? In Settings → General → Send messages with, choose **Enter sends** (Shift+Enter starts a new line) or **⌘Enter sends** (**Ctrl+Enter** on Windows and Linux; Enter starts a new line). ## How do I stop an agent quickly? Press `Esc` while the message box is focused and empty. A message you send while the agent is working waits in a queue above the message box; choose **Send now** on it to interrupt the turn and send it straight away. ## Where to go next - [Getting started](https://micromanager.win/docs/getting-started) - [Workspaces](https://micromanager.win/docs/workspaces) --- title: "Troubleshooting microManager" description: "Fixes for agents that are not found, pull request buttons that fail, setup scripts, ports, missing .env files, Conductor import and chat history." url: "https://micromanager.win/docs/troubleshooting" updated: "2026-11-08" --- # Troubleshooting Most problems come from something microManager cannot see from your login shell, such as an agent CLI not on the PATH, the GitHub CLI signed out, or a file that is tracked rather than ignored. Each section below names the symptom and the fix. Updated 8 November 2026 ## Why does an agent show as not found? On macOS and Linux, microManager looks commands up on your **login shell’s** `PATH`, the same one a new terminal window gets. Check what that shell finds (use `bash -lc` if your shell is bash): ```sh zsh -lc 'command -v claude codex grok opencode gemini' ``` On Windows it uses the `PATH` Windows gives apps, including npm’s `.cmd` shims. Check it in a new PowerShell window: ```powershell where.exe claude codex grok opencode gemini ``` If a command is missing on Windows, reinstall the agent or add its folder to your user `PATH` in Windows’ own settings, then restart microManager. On macOS and Linux, if the command is missing from the login shell but works in your terminal, it is probably added to `PATH` in a file only interactive shells read (such as `.zshrc`). Either move that line to `.zprofile`, or open Settings → Agents and set the profile’s command to the absolute path. **Detect** checks it again and shows the version it found; **Not working** instead of **Not found** means the command is there but its version check failed, and hovering over it shows why. ## Why is Codex found but not starting? If `codex` is the npm launcher, microManager looks for the native binary the package installs beside it. If you installed Codex some other way, set the Codex profile’s command to the native `codex` binary’s absolute path. ## Why do Create PR, Checks or Merge do nothing? Pull request features use the GitHub CLI. Make sure it is installed and signed in: ```sh gh auth status ``` If it is not signed in, run `gh auth login`, then give it a minute: microManager checks the open workspace’s pull request every minute and other workspaces’ pull requests every five minutes. ## What do I do when the setup script fails? Choose **Show output** on the bar above the chat, or open the **Setup** tab of the terminal drawer (**⌘J**, or Ctrl+J on Windows and Linux), to read the output. Fix the cause, then choose **Run again** on the bar or **Run setup again** in the drawer. Remember that scripts run in the workspace folder, not the repository’s main checkout; use `$MICROMANAGER_ROOT_PATH` to reach files there. ## Why does a dev server say its port is in use? Every workspace has its own ten ports starting at `$MICROMANAGER_PORT` (also `$CONDUCTOR_PORT`). Start servers on that port rather than a fixed one. If a script must use a fixed port, set `scripts.run_mode = "nonconcurrent"` so only one workspace runs it at a time. ## Why was my `.env` not copied? Only files git ignores or does not track are copied, and a file that already exists in the new worktree is never overwritten. Also check which list is in effect: a `.worktreeinclude` file in the repository root replaces `file_include_globs`. See [files to copy](https://micromanager.win/docs/files-to-copy). ## Why do my scripts run in PowerShell on Windows? microManager runs setup, run and archive scripts in Git Bash, which comes with Git for Windows, and in PowerShell only when it cannot find Git Bash. Install [Git for Windows](https://git-scm.com/downloads/win), then restart microManager, so bash scripts written on a Mac work as they are. Claude Code then runs its own shell commands in Git Bash too, instead of PowerShell. ## Why does Windows warn that the installer is from an unknown publisher? The Windows installer is not code-signed yet, so Microsoft Defender SmartScreen may warn about it. Check the file’s SHA-512 against `latest.yml` as the [download page](https://micromanager.win/install#windows) shows, then choose **More info** and **Run anyway**. ## Why did Import from Conductor skip some workspaces? Import from Conductor is only offered on a Mac, because that is where Conductor keeps its data. A repository is skipped when Conductor hides it or its folder is gone (the preview marks it **Folder missing**), and an active workspace when its folder no longer exists. Archived workspaces are skipped unless you switch on **Include archived workspaces**; they come over archived, ready to restore, even though their folders are gone. Chats with agents microManager does not run, such as Cursor, are skipped too. Running the import again is safe: anything already imported is skipped, so it only adds what is missing. ## Why does an imported chat have no earlier messages? Earlier messages are read from the agent’s own transcript files: Claude Code’s under `~/.claude/projects/` and Codex’s under `~/.codex/sessions/`, including the older ones Codex compresses to `.jsonl.zst`. If those files were deleted, the chat shows no earlier messages and the agent no longer has the conversation either, so your next message starts a new one with it, and the chat tells you so. Chats with ACP agents such as OpenCode have no transcript microManager can read, so they show no earlier messages. Very long chats keep their newest 2,000 events. ## Why was the branch not renamed after my first message? Automatic naming only renames a branch that still has its placeholder city name, has never been pushed and has no pull request, and only when **Name workspaces from the first message** is on in Settings → General. Rename it yourself from the workspace header if you need to. ## What if nothing here helps? Email with what you did, what you expected and what happened, plus your microManager version and your operating system and its version. For security problems, write to . ## Where to go next - [Agents and profiles](https://micromanager.win/docs/agents) - [Scripts](https://micromanager.win/docs/scripts) - [FAQ](https://micromanager.win/docs/faq) --- title: "microManager FAQ" description: "Short answers about microManager's price, privacy, supported systems and agents, Conductor compatibility, updates and how it relates to Conductor." url: "https://micromanager.win/docs/faq" updated: "2026-11-09" --- # Frequently asked questions microManager is free, needs no account, runs on macOS, Windows and Linux with the agent CLIs you already use, and reads Conductor’s settings files so you can switch the same day. The answers below cover price, privacy, compatibility and updates. Updated 9 November 2026 ## Where do I start? New to microManager? Read [getting started](https://micromanager.win/docs/getting-started). Coming from Conductor? The [migration guide](https://micromanager.win/migrate-from-conductor) walks through the import, and the [shutdown timeline](https://micromanager.win/conductor-shutdown) lists the dates that matter. Something not working? Try [troubleshooting](https://micromanager.win/docs/troubleshooting). ## Where to go next - [Getting started](https://micromanager.win/docs/getting-started) - [Troubleshooting](https://micromanager.win/docs/troubleshooting) - [Privacy](https://micromanager.win/privacy) ## Frequently asked questions ### Is microManager free? Yes. microManager is free to download and use, with no account and no paid tier. You pay your agent providers (Anthropic, OpenAI, SpaceXSI and so on) directly, the same way you do when you use their CLIs on their own. ### Do I need an account? No. There is no microManager account or sign-in. Your agent CLIs keep their own sign-ins, and the GitHub CLI keeps its own for pull requests. ### Does microManager send my code or chats anywhere? No. Workspaces, chats and settings stay on your computer, and microManager has no telemetry. Your agents talk to their own providers exactly as they do outside microManager, and git and the GitHub CLI talk to your remotes. Beyond that, microManager only contacts micromanager.win, to check for updates. The one exception is naming. When Claude Code is installed, the first message of each new chat goes through it to Anthropic’s Haiku model, which names the chat and a new workspace, whichever agent the chat runs. Turn off “Name workspaces and chats from the first message” in Settings → General to name them on your computer instead. ### Which systems does it run on? macOS 13 Ventura or later on Apple Silicon (and on Intel for releases that have an Intel build), Windows on x64 PCs, and Linux on x64, as an AppImage or a .deb. Windows on ARM and ARM64 Linux are not supported yet. ### Which coding agents does it support? Claude Code, Codex, Grok Build, OpenCode and Gemini CLI are built in, and you can add any other agent that speaks the Agent Client Protocol (ACP). ### Does microManager need its own API keys? No. microManager runs the Claude Code and Codex CLIs you already have installed and signed in, so they work the way they do in your terminal, with whatever account each one is signed in to. It does not sign you in to anything itself. ### Can I import my Conductor workspaces? Yes, on a Mac, where Conductor keeps its data. Import from Conductor reads Conductor's local database read-only and brings over your repositories, workspaces, chat tabs, agent paths and settings in one step, leaving every worktree where it is. It never writes to Conductor's data. ### Will my .conductor/settings.toml and scripts keep working? Yes. microManager reads .conductor/settings.toml, .conductor/settings.local.toml, conductor.json and .worktreeinclude, and sets the CONDUCTOR\_\* environment variables Conductor gave local scripts, with the same meanings, alongside its own MICROMANAGER\_\* ones. Imported workspaces get new ports. ### Can I use microManager and Conductor side by side? Yes, while Conductor's local app still works. Importing leaves Conductor's data and your worktrees untouched, so both apps can show the same workspaces. Avoid running agents in the same worktree from both apps at once. ### How does microManager update? It checks micromanager.win when it starts and every six hours, downloads new versions in the background and shows a Restart to update button when one is ready. ### Is microManager open source? Not at the moment. It is free to use; whether to publish the source has not been decided. ### Is microManager made by Conductor or Meta? No. microManager is independent and is not affiliated with, endorsed by or sponsored by Conductor, its maker or Meta. --- title: "How to run multiple coding agents at once | microManager" description: "Run Claude Code, Codex, Grok Build, OpenCode and Gemini CLI side by side on macOS, Windows or Linux, each task in its own workspace and git worktree." url: "https://micromanager.win/features/parallel-agents" updated: "2026-11-08" --- # Run several coding agents at once microManager lets you run many agent sessions at the same time, across repositories, with each task in its own workspace and git worktree so agents never edit the same files. You see at a glance which agents are working, which need an answer and which are done. Updated 8 November 2026 ![A Codex chat on GPT-6-Luna ending with all 12 tests passing after the npm test runs it was allowed once outside its sandbox, above the terminal drawer with Setup, Run server (still running) and Run tests tabs, Run tests showing each passing test and the totals.](https://micromanager.win/screenshots/codex-chat-run-scripts.png) Codex in its own workspace, with the server and the tests started as run scripts in the terminal drawer. ## Why run agents in parallel? A coding agent spends most of a turn working on its own: reading code, running tests, editing files. While one agent works on a bug fix, another can add a feature and a third can write tests. Running them side by side turns that waiting time into progress on other tasks, and you spend your own time reviewing and steering instead of watching. The catch is isolation. Two agents editing the same checkout overwrite each other’s files and confuse each other’s test runs. microManager gives each task its own [git worktree](https://micromanager.win/features/git-worktrees), so agents work in separate folders on separate branches of the same repository. ## How do I keep track of many agents? The sidebar groups workspaces by repository and marks each one: - a spinner while an agent is working, - a mark when it is waiting for your answer or permission, - an error mark when a turn failed, - a dot for replies you have not read, and - a check when you have marked the workspace Done or its pull request is merged. It also shows each workspace’s branch, how many lines it has changed and its pull request state. Pin the workspaces you care about most; the rest sort by latest activity. **⌘1** to **⌘9** and the **⌘K** command palette jump between them (Ctrl instead of ⌘ on Windows and Linux). ## Will I miss it when an agent finishes? No. When a turn ends or an agent needs input, and the window is in the background or you are looking at another workspace, microManager sends a system notification, with an optional sound. The app’s icon shows how many workspaces need you: a count on the Dock icon on a Mac and in Linux launchers that show one, and a dot on the flashing taskbar button on Windows. **Keep the computer awake**, on by default, stops your computer from sleeping while any agent is working, so long jobs are not paused. ## Can one workspace have several chats? Yes. Press **⌘T** (Ctrl+T on Windows and Linux) for a new tab. Each tab is its own agent session with its own agent, model, effort level and permission mode, all working in the same worktree. Use one tab to implement and another to review, or give a second agent a narrower job without cluttering the first chat. ## What does each chat show? Messages and replies with formatted code, the agent’s thinking (collapsed), a card for each tool call (commands with their output, file edits as small diffs, searches, subagents and MCP calls), the agent’s to-do list while a turn runs, and cards for permission requests and questions. Each turn ends with its duration, tokens and cost when the agent reports them. ## Can I talk to an agent while it works? Yes. With Claude Code and Codex, a message you send during a turn is added to that turn: the agent reads it at its next step and carries on with it, so you can correct course without stopping the work. **Alt+Enter** queues the message for when the turn ends instead, and Settings → General can make queueing the default. Other agents always get it queued. You can edit or delete queued messages, add one to the running turn later, or choose **Send now** to interrupt the turn and send immediately. **Esc** in an empty message box stops the running turn. ## Where to go next - [How workspaces work](https://micromanager.win/docs/workspaces) - [Supported agents](https://micromanager.win/agents) ## Frequently asked questions ### How many agents can I run at once? microManager sets no limit. The practical limits are your computer's memory and CPU, and the rate limits of your agent providers. ### Can I mix different agents in one workspace? Yes. Each workspace can have several chat tabs, and each tab picks its own agent, so a Codex tab can review what a Claude Code tab wrote in the same worktree. ### What happens when an agent needs me while I am in another workspace? The workspace shows that it needs input in the sidebar, and a system notification with an optional sound appears when the window is in the background or another workspace is open. On a Mac the Dock icon counts the workspaces waiting for you or with unread replies, as do Linux launchers that show counts; on Windows the taskbar button gets a dot and flashes. --- title: "A git worktree per coding agent | microManager" description: "Every microManager workspace is a real git worktree on its own branch, with your ignored files copied in, ten ports reserved and your setup script run." url: "https://micromanager.win/features/git-worktrees" updated: "2026-11-08" --- # One git worktree per task Every microManager workspace is a real git worktree on its own branch, so each agent edits its own copy of the repository while sharing one .git folder. microManager creates it, copies in your ignored files, reserves ports and runs your setup script, then removes it when you archive the workspace. Updated 8 November 2026 ![The New workspace dialog over the app: repository ledger-api, a new branch from main, a name to be set from the first message, agent Claude Code, model Haiku, effort Agent default, mode Accept edits, and a first message asking it to reject account names longer than 60 characters.](https://micromanager.win/screenshots/new-workspace-dialog.png) A new workspace is a git worktree on its own branch, with an agent ready to work in it. ## Why does each agent need its own worktree? Agents edit files, run builds and start servers. Two of them in one checkout step on each other: one agent’s half-finished edit breaks the other’s tests, and both try to start a dev server on the same port. Separate clones avoid that but are slow to make and keep separate copies of the history. A git worktree is the middle path: a separate folder and branch per task, sharing one `.git` object store. ## What does microManager set up for each worktree? | Step | What happens | | ------------- | ----------------------------------------------------------------------------------------------------------------------------------------- | | Branch | A new branch from the base you pick, named after a city (`/lisbon`) and renamed after your first message unless you turn that off | | Worktree | `git worktree add` into `~/micromanager/workspaces//` | | Ignored files | `.env` and other ignored files copied in from your main checkout | | Ports | Ten ports reserved, the first in `MICROMANAGER_PORT` | | Setup | Your setup script runs in a terminal you can watch | | Shared notes | A `.context/` folder git ignores, for notes agents share | Details are in the [workspaces docs](https://micromanager.win/docs/workspaces). ## How are ignored files handled? A fresh worktree only has committed files. microManager copies the rest from your main checkout using `.worktreeinclude`, or `file_include_globs` from your repository settings, or `.env*` by default, and it never overwrites a file that already exists. See [files to copy](https://micromanager.win/docs/files-to-copy). ## How do ports work when several dev servers run? Each workspace gets ten ports of its own, starting at a multiple of 10 from 51000 up. Start servers on `$MICROMANAGER_PORT` (or `$CONDUCTOR_PORT` in scripts written for Conductor) and every workspace can run its dev server at the same time. See [environment variables](https://micromanager.win/docs/environment-variables). ## What happens to the worktree when I am done? Archiving runs your archive script, saves any uncommitted work as a hidden git commit, removes the worktree folder and keeps the branch, so the disk space comes back and only ignored files such as `node_modules` are gone. You can restore an archived workspace later, at the same path, on the same branch, with its uncommitted work back in place. ## Does this work with the worktrees Conductor made? Yes. Importing from Conductor registers each existing worktree where it already is, on its existing branch. Nothing is moved, copied or rewritten. ## Where to go next - [Workspaces in detail](https://micromanager.win/docs/workspaces) - [Files to copy](https://micromanager.win/docs/files-to-copy) - [Environment variables and ports](https://micromanager.win/docs/environment-variables) ## Frequently asked questions ### What is a git worktree? A git worktree is an extra working folder attached to the same repository. Each worktree has its own branch and files, but they all share one object store, so a new worktree is quick to create and costs little disk space compared with a second clone. ### Are worktrees a security boundary? No. Worktrees keep agents from editing the same files, but every agent still runs as your user on your computer. Use each agent's own permission settings to limit what it may do. ### Can I open a workspace in my editor? Yes. The Open in menu opens the worktree in VS Code, Cursor, Zed or Windsurf on every system, plus Finder, Terminal, iTerm or Xcode on a Mac, File Explorer or Windows Terminal on Windows, and your file manager or terminal on Linux. It can also open the branch on GitHub. --- title: "Review agent diffs and send comments back | microManager" description: "See every change an agent made, comment on lines in a unified or split diff, and send the comments back so the agent fixes them. Agents can comment too." url: "https://micromanager.win/features/diff-review" updated: "2026-11-08" --- # Review diffs, comment, send back microManager shows every change in a workspace as a diff against the branch you will merge into, untracked files included. Click a line number to comment, then send your comments to the agent, which reads them and makes the fixes in the same chat. Updated 8 November 2026 ![The diff of src/app.ts in the Branch view, with the new GET /transfers route highlighted and a line comment under it asking for a ?limit= parameter, a Send 1 to agent button, and the Changes panel listing accounts.ts, app.ts and app.test.ts with the lines each added.](https://micromanager.win/screenshots/diff-review-comments.png) Comment on lines of a diff, then send the comments to the agent. ## What does the Changes panel show? Every file the workspace changed, as a tree, each with its status (added, modified, deleted, renamed) and its added and removed line counts. A toggle switches between two views: | View | Compares | Good for | | ----------- | ------------------------------------------------------------------------------------- | ---------------------------------------------- | | Branch | The merge base with the target branch, against the working tree, plus untracked files | Reviewing the whole task before a pull request | | Uncommitted | The last commit against the working tree | Checking what the agent did in the last turn | **⌘⇧E** (Ctrl+Shift+E on Windows and Linux) shows or hides the panel. ## How do I read a diff? Click a file to open its diff over the chat, or beside it. Choose unified or split view; both have syntax highlighting. ## How do comments get back to the agent? Click a line number to comment on that line, or Shift-click or drag across line numbers to comment on a range. Comments collect in the **Review** panel. When you are done, **Send to agent** (it counts them: **Send 2 to agent**) sends every open comment to the chat tab on screen as one message, each with its file and lines, and the agent works through them. Sent comments are marked **Sent**, and you can resolve them once fixed, or reopen them. ## Can agents leave comments too? Yes. Through microManager’s MCP server, an agent can call `GetWorkspaceDiff` to read the workspace’s changes, `DiffComment` to leave comments on specific lines and `GetDiffComments` to read the comments not resolved yet, so “address my comments” works without sending them. Ask one agent to review what another wrote, and its findings land in the same Review panel as yours. See [MCP tools](https://micromanager.win/docs/mcp-tools). ## What if I do not like what the agent did? Roll the workspace back to before the message that caused it with a [checkpoint](https://micromanager.win/features/checkpoints), or tell the agent what to change. When the diff looks right, open a [pull request](https://micromanager.win/features/pull-requests). ## Where to go next - [Checkpoints](https://micromanager.win/features/checkpoints) - [Pull requests](https://micromanager.win/features/pull-requests) - [MCP tools for agents](https://micromanager.win/docs/mcp-tools) ## Frequently asked questions ### Does the diff include files the agent has not committed? Yes. The branch view compares the merge base with the target branch against the working tree, so committed, uncommitted and untracked files all appear. The uncommitted view shows only what changed since the last commit. ### Can an agent review another agent's work? Yes. Agents can read the diff with the GetWorkspaceDiff tool, leave line comments with the DiffComment tool and read yours with the GetDiffComments tool. Their comments appear in the Review panel next to yours. ### Is there a limit on file size? Each file's diff is capped at 1 MB and marked as truncated beyond that, so one huge generated file cannot freeze the view. Binary files are listed without a diff. --- title: "Checkpoints to undo an agent's turn | microManager" description: "microManager snapshots the worktree before every message you send, so you can roll files and the chat back to that point if an agent's turn goes wrong." url: "https://micromanager.win/features/checkpoints" updated: "2026-11-08" --- # Checkpoints before every turn Before each message you send, microManager saves a snapshot of the whole worktree, including uncommitted and untracked files, as a hidden git commit. If a turn goes wrong, choose Revert to here on that message and the files go back to exactly how they were. Updated 8 November 2026 ## How does a checkpoint work? Just before your message goes to the agent, microManager records the worktree’s state as a git commit object, using a temporary index so your own staging area is untouched: 1. it adds every tracked and untracked file to the temporary index, leaving out ignored files and the shared `.context/` folder, 2. writes it as a tree and a commit whose parent is your current `HEAD`, and 3. stores it under `refs/micromanager/checkpoints//`. Your branch, your staged changes and your working files do not change. ## How do I roll back? Each of your messages in the chat has **Revert to here**, which shows when you point at the message. It is available once the running turn ends, and refused while another chat of the workspace is running a turn. After you confirm, microManager: 1. saves the worktree as it is now under another hidden ref, shown in the chat as **Files from before** with a button to copy it, so the revert itself can be undone with git, 2. puts the snapshot’s files back in the worktree, 3. removes untracked files that were not in the snapshot (ignored files and `.context/` stay), and 4. moves the branch back to the commit it was on at that moment. The chat’s later messages are marked as reverted, so it is clear which part of the conversation no longer matches the files. Claude Code sessions are also rewound to that point when the installed Claude Code supports it, so the agent continues from the same state as the files. ## When should I use a checkpoint instead of asking the agent to undo? When the agent changed many files, deleted something, or went down a wrong path for several turns. Asking an agent to undo its own work is slow and often incomplete; a checkpoint puts back exactly what was there, byte for byte. ## Do checkpoints work with every agent? Yes. Snapshots are made by microManager with git, before the message is sent, so they work the same for Claude Code, Codex and ACP agents. Only the chat rewind described above depends on the agent. ## Where to go next - [Diff review](https://micromanager.win/features/diff-review) - [Workspaces](https://micromanager.win/docs/workspaces) ## Frequently asked questions ### Do checkpoints create commits on my branch? No. A checkpoint is a commit object stored under refs/micromanager/checkpoints, outside your branches. It never appears in git log for your branch and is never pushed. ### Are ignored files included? No. Like git itself, a checkpoint leaves out files your .gitignore excludes, such as node_modules or .env, and reverting leaves them alone. ### How long are checkpoints kept? Until the workspace is archived. Archiving deletes that workspace's checkpoint refs. --- title: "Create, check and merge PRs from microManager" description: "microManager creates pull requests through your agent and the GitHub CLI, shows CI checks and review state, and merges with squash, merge or rebase." url: "https://micromanager.win/features/pull-requests" updated: "2026-11-08" --- # Pull requests without leaving the app When a workspace is ready, Create PR asks the agent to commit, push and open a pull request with the GitHub CLI against the workspace’s target branch. microManager then shows the pull request’s checks, review state and mergeability, and merges it for you. Updated 8 November 2026 ![A Grok chat that committed its work as Add balanceText to each account on GET /accounts, a terminal showing that commit on accounts-balance-text with no upstream, and the Checks panel saying the branch was never pushed and that the GitHub CLI is not signed in, with gh auth login to copy.](https://micromanager.win/screenshots/pull-request-checks.png) Before there is a pull request, the Checks panel shows what the branch still needs: here a first push and a signed-in GitHub CLI. ## How does Create PR work? **Create PR** sends the agent in the current chat a prompt asking it to commit the work, push the branch and run `gh pr create` against the workspace’s target branch. Because the agent wrote the code, it can write a meaningful title and description. The prompt is yours to change in Settings → Prompts. ## How do I follow CI and reviews? Once a pull request exists for the workspace’s branch, the header’s pull request button shows its number and state, and the **Checks** panel lists every check with its status and a link, plus whether the pull request can be merged and its review decision. microManager refreshes this every minute for the workspace you have open and every five minutes for the others, and the sidebar badge shows its number, colored by its checks while it is open, or Draft, Merged or Closed. When checks fail, ask the agent to fix them; the fix-errors prompt in Settings → Prompts is there for that. ## How do I merge? **Merge…**, in the pull request button’s menu, runs `gh pr merge` once you confirm the method: **Squash and merge**, **Create a merge commit** or **Rebase and merge**, starting from the one set in Settings → General → Merge method. A draft has to be marked ready for review first. Turn on **Archive after merging** in Settings → General to archive the workspace as soon as a merge from microManager goes through, which removes the worktree and keeps the branch and any uncommitted work. ## Which branch does the pull request target? The workspace’s target branch, shown as **→ target** in the header. It starts as the repository’s default branch; pick another before creating the pull request if the work belongs elsewhere. ## Where to go next - [Diff review](https://micromanager.win/features/diff-review) - [Workspaces and archiving](https://micromanager.win/docs/workspaces) ## Frequently asked questions ### Do I need the GitHub CLI? Yes, for pull request features. Install gh and run gh auth login once; microManager uses whichever account gh is signed in to. Everything else in microManager works without it. ### Can I change how the agent writes pull requests? Yes. The create-PR prompt is editable in Settings → Prompts, alongside the prompts for code review, fixing failing checks, resolving conflicts and renaming branches, and a repository can have its own in Settings → Repositories. ### Does it work with GitLab or Bitbucket? No. Pull request features use the GitHub CLI, so they work with GitHub repositories. Git itself works with any remote. --- title: "ACP agent client for Mac, Windows and Linux | microManager" description: "microManager is an ACP client for Mac, Windows and Linux: Grok Build, OpenCode, Gemini CLI and any ACP agent run in parallel beside Claude Code and Codex." url: "https://micromanager.win/features/acp-agents" updated: "2026-11-08" --- # Any ACP agent, natively microManager is an Agent Client Protocol (ACP) client, so any agent that speaks ACP over stdio runs in it like a built-in one, with chat, permissions, plans and diff review. Grok Build, OpenCode and Gemini CLI come preconfigured, and you can add others in Settings. Updated 8 November 2026 ![Settings, Agents: the default agent set to Claude Code, an empty Favorites list with Add favorite, and the profiles Claude Code, Codex and Grok found with their versions and OpenCode and Gemini CLI not found, with Detect all and Add ACP agent buttons.](https://micromanager.win/screenshots/settings-agents.png) Settings → Agents shows which agent CLIs microManager found and their versions. ## Which ACP agents are built in? | Agent | Command microManager runs | | -------------------------------------------------------- | ------------------------- | | [Grok Build](https://micromanager.win/agents/grok) | `grok agent stdio` | | [OpenCode](https://micromanager.win/agents/opencode) | `opencode acp` | | [Gemini CLI](https://micromanager.win/agents/gemini-cli) | `gemini --acp` | Each runs the CLI you installed and signed in to. If a command is somewhere unusual, set its absolute path in Settings → Agents. ## What works with ACP agents? - **Chat**: streamed replies and the agent’s thinking. - **Tool calls**: shown as cards, with file edits as diffs. - **Plans**: the agent’s plan appears as a to-do list above the message box while it works. - **Permissions**: when the agent asks before acting, a card offers the agent’s own choices. Whether it asks is up to the agent’s own settings; microManager answers for you in **Bypass permissions**, and for file edits in **Accept edits**. - **Models**: picked at the bottom of the message box, when the agent lists them over ACP. - **Modes**: an agent whose own modes have the ids `default`, `acceptEdits`, `plan` and `bypassPermissions`, as Claude Code’s ACP adapter does, switches with the mode picker. Plan is offered only when the agent has a `plan` mode, and Auto never is. - **Slash commands**: the agent’s own commands pass straight through. - **Resuming**: when the agent supports loading sessions, a chat continues the same agent session after a restart. - **microManager’s MCP tools**: every ACP agent gets the micromanager server, over HTTP when the agent supports it or through a stdio bridge when it does not, so it can ask you questions, read the workspace diff and leave review comments. Everything else, from worktrees and checkpoints to diff review and pull requests, is the same for every agent. ## How do I add another ACP agent? In Settings → Agents, choose **Add ACP agent**. Enter a name, the command that starts the agent in ACP mode and any arguments, then add environment variables or models in its profile if it needs them. It joins the agent picker for new workspaces and chats. Google’s Antigravity CLI, for example, is the command `agy` with the argument `acp`. ## What can an ACP agent touch on my computer? Through the protocol, microManager lets an agent read and write text files only inside its workspace, and does not lend it a terminal. The agent process itself still runs as your user with its own tools and permission rules, as it would in your terminal. ## Where to go next - [Agents and profiles](https://micromanager.win/docs/agents) - [Grok Build in microManager](https://micromanager.win/agents/grok) - [OpenCode in microManager](https://micromanager.win/agents/opencode) - [Gemini CLI in microManager](https://micromanager.win/agents/gemini-cli) ## Frequently asked questions ### What is the Agent Client Protocol? ACP is an open protocol, based on JSON-RPC, for connecting coding agents to the apps that host them. An app that speaks it can run any agent that implements it without custom integration work. ### Do I need an adapter or shim for Grok Build? No. microManager runs grok agent stdio directly and talks to it over ACP. ### Can an ACP agent read and write files? microManager lets ACP agents read and write text files only inside their own workspace. It does not offer them its terminal capability; agents run commands with their own tools, under their own permission prompts. --- title: "Run multiple Claude Code agents in parallel | microManager" description: "Run several Claude Code agents at once, each in its own git worktree, with the Claude Code you have installed and signed in. Free, with diffs and PRs." url: "https://micromanager.win/agents/claude-code" updated: "2026-11-08" --- # Claude Code in microManager microManager runs the Claude Code you already have installed and signed in, through the Claude Agent SDK, so every chat is a full Claude Code session with your CLAUDE.md, settings, hooks and MCP servers, and the repository’s own once you trust it. You can run as many as you like at once, each in its own git worktree. Updated 8 November 2026 ![A microManager window while Claude Code, on Haiku, works in the workspace Reject overlong account names: edit cards showing its changes to test/accounts.test.ts and test/app.test.ts as diffs, a Run the test suite card that was allowed once, and its reply still streaming above a Working… line. The sidebar lists two more ledger-api workspaces, and the Changes panel lists three changed files.](https://micromanager.win/screenshots/workspace-claude-chat.png) Claude Code in the middle of a turn: each edit shows as a diff, the test run it asked for was allowed once, and its reply is still streaming. Codex and Grok each work in another workspace. ## How do I run several Claude Code agents at once? Create a workspace for each task (**⌘N**, or Ctrl+N on Windows and Linux) and pick Claude Code. Each workspace is its own git worktree on its own branch, so the sessions never edit the same files, and each can have more chat tabs (**⌘T**) for a second Claude working beside the first. The sidebar shows which sessions are working, which need an answer and which are done, and a notification tells you when one finishes. See [parallel agents](https://micromanager.win/features/parallel-agents). ## How does microManager run Claude Code? Each chat tab is a Claude Code session started through the Claude Agent SDK, with the SDK pointed at **your** `claude` executable: the one on your login shell’s `PATH`, or a wrapper script you choose. The session runs in the workspace’s worktree with the workspace’s environment, loads your user settings, and the repository’s project and local settings once you [trust the repository](https://micromanager.win/docs/scripts#why-did-a-script-from-the-repository-not-run), and keeps Claude Code’s own system prompt. microManager appends a short note about the workspace: its name and folder, the repository’s main checkout, the branch and its target branch, the shared `.context/` folder, the reserved ports and its own tools, then your general prompt from Settings → Prompts. ## Which models and effort levels can I use? The model picker offers `opus`, `sonnet`, `haiku` and `fable`, and accepts any model id you type, so new models work the day your Claude Code supports them. Effort goes from low through medium, high and xhigh to max. Both can change per chat. ## How do permissions and questions show up? Pick a permission mode per chat: **Ask first**, **Accept edits**, **Plan**, **Auto** (Claude Code’s own reviewer model decides each request) or **Bypass permissions**. New chats start in the `permissions.defaultMode` of your Claude Code settings, or in the mode you set for Claude Code in Settings → Agents. When Claude Code asks to use a tool, a card shows what it wants to do, with **Allow**, **Deny** and, when Claude Code suggests one, an always-allow choice such as **Allow all edits this session**. At the end of planning, **Approve plan** or **Keep planning** answers its plan. When it uses its `AskUserQuestion` tool, you get a card with its options (single or multiple choice, or your own text), and your answer goes straight back to it. ## What else does a Claude Code chat show? Streamed replies with formatted code, its thinking (collapsed), a card per tool call (commands with output, edits as diffs, searches, subagents), its to-do list while it works, and each turn’s duration, tokens and cost. **Esc** stops a turn; checkpoints before every message let you roll a turn back, and the session rewinds with it when your Claude Code version supports that. ## Can I use a wrapper script instead of claude? Yes. Set the Claude Code profile’s command to any executable that behaves like `claude`, such as a script that sets a proxy or chooses an account. If Conductor was set up with a custom Claude Code path, the import uses it. ## What tools does Claude get from microManager? The `micromanager` MCP server runs inside the session, so Claude can call `GetWorkspaceDiff`, `DiffComment`, `GetDiffComments`, `GetTerminalOutput` and `GetWorkspaceInfo`. For questions it uses its own built-in `AskUserQuestion`. See [MCP tools](https://micromanager.win/docs/mcp-tools). ## What do I need? Claude Code installed and signed in, so that `claude` works in your terminal. On Windows, Claude Code runs its shell commands in Git Bash when Git for Windows is installed, and in PowerShell otherwise. Then add a repository, press **⌘N** (Ctrl+N on Windows and Linux) and pick Claude Code. ## Where to go next - [Agents and profiles](https://micromanager.win/docs/agents) - [Codex in microManager](https://micromanager.win/agents/codex) - [Run agents in parallel](https://micromanager.win/features/parallel-agents) ## Frequently asked questions ### Do I need an Anthropic API key for microManager? No. microManager starts your claude CLI, which uses whatever sign-in it already has. microManager itself has no account and never sees your credentials. ### Will my CLAUDE.md, hooks and MCP servers apply? Yes. Sessions always load your own Claude Code settings. The repository’s project and local settings (its CLAUDE.md, .claude hooks and permission rules, and .mcp.json) load once you trust the repository, which repositories imported from Conductor already are; then everything behaves as it does in your terminal. ### Can I continue a Claude Code session I started in Conductor? Yes. Importing from Conductor brings each chat over with its Claude session id, loads its history from Claude Code's own transcript file, and the next message continues the same session. --- title: "Run multiple Codex agents in parallel | microManager" description: "microManager runs your installed Codex CLI through codex app-server, several threads at once, each in its own git worktree, with approvals, diffs and PRs." url: "https://micromanager.win/agents/codex" updated: "2026-11-08" --- # Codex in microManager microManager runs the Codex CLI you already have installed and signed in, through Codex’s app-server protocol, so each chat is a Codex thread with streamed replies, approvals and model choice. Run several at once, each in its own git worktree, next to Claude Code or any other agent. Updated 8 November 2026 ![A Codex chat on GPT-6-Luna ending with all 12 tests passing after the npm test runs it was allowed once outside its sandbox, above the terminal drawer with Setup, Run server (still running) and Run tests tabs, Run tests showing each passing test and the totals.](https://micromanager.win/screenshots/codex-chat-run-scripts.png) Codex in its own workspace, with the server and the tests started as run scripts in the terminal drawer. ## How do I run several Codex agents in parallel? Create one workspace per task and pick Codex for each. Every workspace is its own git worktree and branch, so the Codex threads never edit the same files, and a workspace can hold more than one Codex chat tab. You review each one’s diff on its own and open a pull request per branch. Claude Code, Grok Build or any other agent can work in the next workspace over. See [parallel agents](https://micromanager.win/features/parallel-agents). ## How does microManager run Codex? Each chat starts `codex app-server` and talks to it over standard input and output with JSON-RPC, the protocol Codex provides for apps that host it. A chat is a Codex thread: microManager starts or resumes the thread, sends your messages as turns, streams the replies and tool activity into the chat, and can interrupt a turn. If the app-server cannot be used, microManager falls back to `codex exec --json`. ## Which codex does it run? The one on your login shell’s `PATH`, or the path you set in Settings → Agents. When `codex` is the npm launcher script, microManager runs the native Codex binary the package installs beside it. If you set a custom Codex path in Conductor, the import carries it over. ## How do approvals work? When Codex asks before running a command or applying a change, a card in the chat shows the request with **Allow once**, **Allow for this session** and **Reject**. The mode picker sets Codex’s sandbox: **Ask first** lets it write in the workspace and asks for anything more, **Accept edits** also approves its file changes for you, **Plan** keeps it read-only, **Auto** lets Codex’s own reviewer answer its requests, and **Bypass permissions** gives it full access with no approvals. ## Which models and reasoning levels can I use? The model picker lists the models Codex has cached in `~/.codex/models_cache.json` from the moment the app starts, refreshed from the app-server once a Codex chat runs, and takes any model id you type. The model and effort in your `~/.codex/config.toml` become the profile’s defaults when you have not chosen others. Effort goes from low through medium, high, xhigh and max to ultra, and the picker offers the levels the chosen model supports. ## What tools does Codex get from microManager? microManager adds its `micromanager` MCP server to the Codex session through Codex’s own configuration flags, connected through a small stdio bridge to a local endpoint that only that session’s token can use. Codex can then ask you questions, read the workspace diff, leave review comments and read terminal output. See [MCP tools](https://micromanager.win/docs/mcp-tools). ## Is my Codex history kept? Codex keeps each thread in its own session files under `~/.codex/sessions`, compressing older ones to `.jsonl.zst`, which microManager reads too. When microManager imports a chat that has a thread id, it reads that history (up to the newest 2,000 events) and the next message resumes the same thread. ## What do I need? The Codex CLI installed and signed in, so that `codex` works in your terminal. ## Where to go next - [Agents and profiles](https://micromanager.win/docs/agents) - [Claude Code in microManager](https://micromanager.win/agents/claude-code) - [Diff review](https://micromanager.win/features/diff-review) ## Frequently asked questions ### Do I need an OpenAI API key for microManager? No. microManager starts your codex CLI, which uses whatever sign-in it already has. microManager has no account and never sees your credentials. ### Which Codex models can I pick? The models Codex has cached for your account, read when microManager starts and refreshed from Codex's app-server once a Codex chat runs. For any other model, type its id. The model in your Codex configuration is the default. ### Can I continue Codex chats from Conductor? Yes. Importing brings each Codex chat over with its thread id and loads its history from Codex's own session files, so the next message continues the same thread. --- title: "A desktop GUI for Grok Build agents | microManager" description: "microManager is a free desktop GUI for SpaceXSI's Grok Build: it runs grok agent stdio over ACP, in parallel git worktrees beside Claude Code and Codex." url: "https://micromanager.win/agents/grok" updated: "2026-11-09" --- # A desktop GUI for Grok Build microManager is a free desktop GUI for Grok Build, the coding agent of SpaceXSI (formerly xAI), on macOS, Windows and Linux: it starts grok agent stdio and talks to it over the Agent Client Protocol, with no adapter in between. Each chat is a Grok session in its own git worktree, next to Claude Code, Codex or anything else you run. Updated 9 November 2026 ![A Grok chat on Grok 4.7 Fast in the workspace Add balanceText to accounts route: an edit card showing the test it added to test/app.test.ts as a diff, an Execute npm test card, a folded Thought, and its summary ending with npm test: 11 passed, 0 failed.](https://micromanager.win/screenshots/grok-chat.png) Grok Build over ACP: its edits show as diffs, its commands as cards, and its thinking folds away. ## Is there a GUI for Grok Build? Grok Build is a command-line agent. microManager gives it a desktop window: each Grok session is a chat with streamed replies, tool calls shown as cards, file edits as diffs, permission requests as buttons and its plan as a to-do list, and several sessions can run at once in separate worktrees. Because it speaks ACP, the protocol Grok Build supports for hosting apps, there is no wrapper script or shim to maintain. ## How does microManager run Grok Build? The built-in profile, named Grok, runs: ```sh grok agent stdio ``` and talks to it over the Agent Client Protocol on standard input and output. It uses the `grok` you installed and signed in to; if it is not on your login shell’s `PATH`, set its absolute path in Settings → Agents. ## What works with Grok Build? Everything ACP offers: streamed replies and thinking, tool calls shown as cards, its plan as a to-do list while it works, permission requests whenever Grok asks (its own permission setting decides whether it does), model choice when Grok offers several, and its slash commands. Grok also gets microManager’s [MCP tools](https://micromanager.win/docs/mcp-tools), and every workspace feature applies: worktrees, checkpoints, diff review and pull requests. ## What happens to Grok chats when I import from Conductor? ACP chats that Conductor ran with a Grok model become chats on the Grok profile, in the same workspace. If Conductor was set up to reach Grok through a separate harness program, microManager does not carry that path over, because it runs `grok agent stdio` itself; it turns the Grok profile on instead. ## What do I need? Grok Build installed and signed in, so that `grok` works in your terminal. ## Where to go next - [ACP agents](https://micromanager.win/features/acp-agents) - [Agents and profiles](https://micromanager.win/docs/agents) ## Frequently asked questions ### Do I need a shim or adapter? No. microManager starts grok agent stdio and speaks ACP to it directly. ### Are my Grok chats from Conductor imported? Yes. ACP chats that Conductor ran with a Grok model are imported as chats on microManager's Grok profile. --- title: "Run OpenCode in parallel worktrees | microManager" description: "microManager runs OpenCode over the Agent Client Protocol with opencode acp, several sessions at once, each in its own git worktree with diffs and PRs." url: "https://micromanager.win/agents/opencode" updated: "2026-11-08" --- # OpenCode in microManager microManager runs OpenCode over the Agent Client Protocol by starting opencode acp, so you can run OpenCode sessions in parallel, each in its own git worktree, with whatever models and providers you set up in OpenCode. Updated 8 November 2026 ## How does microManager run OpenCode? The built-in OpenCode profile runs: ```sh opencode acp ``` in the workspace’s worktree and talks to it over the Agent Client Protocol. It uses the `opencode` on your login shell’s `PATH`, or the absolute path you set in Settings → Agents. ## What works with OpenCode? Streamed replies, tool calls as cards with file edits shown as diffs, its plan as a to-do list, permission requests, model choice and slash commands, plus microManager’s [MCP tools](https://micromanager.win/docs/mcp-tools). Worktrees, checkpoints, diff review and pull requests work exactly as they do for every other agent. ## What happens to OpenCode chats when I import from Conductor? ACP chats that Conductor ran with a non-Grok model become OpenCode chats in microManager, with the model name carried over. Their earlier messages are not shown, and a chat whose session Conductor kept in its own store, which no ACP agent can load, starts a new OpenCode session. ## What do I need? OpenCode installed and configured with at least one provider, so that `opencode` works in your terminal. ## Where to go next - [ACP agents](https://micromanager.win/features/acp-agents) - [Agents and profiles](https://micromanager.win/docs/agents) ## Frequently asked questions ### Which models can I use with OpenCode? Whatever your OpenCode setup offers. microManager shows the models OpenCode reports over ACP and lets you pick one per chat. ### Do my OpenCode settings apply? Yes. microManager starts your own opencode command, so its configuration, providers and sign-ins are used as they are. --- title: "Run Gemini CLI in parallel worktrees | microManager" description: "microManager runs Google's Gemini CLI over the Agent Client Protocol with gemini --acp, in parallel git worktrees next to your other agents." url: "https://micromanager.win/agents/gemini-cli" updated: "2026-11-08" --- # Gemini CLI in microManager microManager runs Google’s Gemini CLI over the Agent Client Protocol by starting gemini --acp, so Gemini sessions run in parallel with your other agents, each in its own git worktree. Updated 8 November 2026 ## How does microManager run Gemini CLI? The built-in Gemini CLI profile runs: ```sh gemini --acp ``` in the workspace’s worktree and talks to it over the Agent Client Protocol. It uses the `gemini` on your login shell’s `PATH`, or the absolute path you set in Settings → Agents. The profile’s arguments are editable, so you can follow changes to Gemini CLI’s flags without waiting for a microManager update. ## Who can still use Gemini CLI? Google’s [Gemini CLI README](https://github.com/google-gemini/gemini-cli) says that since 18 June 2026 it is for organizations with Gemini Code Assist Standard or Enterprise, or Code Assist for GitHub, on Google Cloud, plus older paid Gemini API and Vertex AI keys. Free Google accounts and Google AI Pro and Ultra moved to [Antigravity CLI](https://antigravity.google/docs/cli/acp) (`agy`). Antigravity CLI speaks the Agent Client Protocol too. To use it, sign in once by running `agy` in a terminal, then in Settings → Agents choose **Add ACP agent** with the command `agy` and the argument `acp`. ## What works with Gemini CLI? Streamed replies, tool calls as cards, plans, permission requests, model choice when Gemini offers it and slash commands, plus microManager’s [MCP tools](https://micromanager.win/docs/mcp-tools). Worktrees, checkpoints, diff review and pull requests work as they do for every agent. ## What do I need? Gemini CLI installed and signed in with an account it still serves, so that `gemini` works in your terminal. ## Where to go next - [ACP agents](https://micromanager.win/features/acp-agents) - [Agents and profiles](https://micromanager.win/docs/agents) ## Frequently asked questions ### Can I use Gemini CLI with a free Google account or Google AI Pro? Not any more. Since 18 June 2026, Gemini CLI is for Gemini Code Assist Standard and Enterprise, Code Assist for GitHub on Google Cloud, and paid Gemini API or Vertex AI keys. Free accounts and Google AI Pro and Ultra moved to Antigravity CLI, which you can add to microManager as an ACP agent with the command agy and the argument acp. ### Does Gemini CLI need anything special? Only to be installed and signed in with an account that can still use it. microManager starts it with its ACP flag; if your version uses a different flag, change the arguments in Settings → Agents. ### Can Gemini review another agent's work? Yes. Open a Gemini tab in the same workspace and ask it to review; it can read the diff and leave line comments through microManager's MCP tools. --- title: "Conductor alternative for Mac, Windows, Linux: microManager" description: "microManager is a free Conductor alternative: parallel coding agents in git worktrees, the same settings.toml and CONDUCTOR_* vars, one-click import." url: "https://micromanager.win/conductor-alternative" updated: "2026-11-08" --- # The Conductor alternative that imports your workspaces microManager is a free Conductor alternative for macOS, Windows and Linux: it runs Claude Code, Codex and other agents in parallel git worktrees with diff review and pull requests, and on a Mac it imports your Conductor repositories, workspaces and chats in one click. Your .conductor/settings.toml scripts and CONDUCTOR\_\* variables keep working, so you can switch the same day. Updated 8 November 2026 [Download](https://micromanager.win/download) [Migration guide](https://micromanager.win/migrate-from-conductor) `curl -fsSL https://micromanager.win/install.sh | sh` ## Why are people leaving Conductor now? Meta bought Conductor on 4 November 2026, and new Mac app downloads stopped that day. The cloud service is shut down from 20 November 2026, and the Mac app winds down in early 2027. [Full timeline with sources](https://micromanager.win/conductor-shutdown). 1. 4 November 2026 Wednesday ### Meta acquires Conductor; sign-ups and Mac downloads stop Sources: [TechCrunch](https://techcrunch.com/2026/11/04/meta-acquires-ai-coding-startup-conductor/), [SiliconANGLE](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), [Conductor](https://www.conductor.build/announcement) 2. 13 November 2026 Friday ### Conductor Cloud goes read-only Sources: [SiliconANGLE](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), [Conductor](https://www.conductor.build/announcement) 3. 20 November 2026 Friday ### Conductor Cloud shuts down Sources: [TechCrunch](https://techcrunch.com/2026/11/04/meta-acquires-ai-coding-startup-conductor/), [SiliconANGLE](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), [Conductor](https://www.conductor.build/announcement) 4. Early 2027 Exact date not announced ### The local Mac app winds down Sources: [SiliconANGLE](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), [Conductor](https://www.conductor.build/announcement) ## What should a Conductor replacement do? The jobs Conductor’s local app did every day, and how microManager does each one. | You need | microManager | | -------------------------------- | ------------------------------------------------------------------------------------------------------------------------------ | | Several agents at once, isolated | Each workspace is its own git worktree and branch, with ten ports and your setup script. | | The agents you already pay for | Runs the Claude Code, Codex, Grok Build, OpenCode and Gemini CLI you installed and signed in, plus any ACP agent. | | Your existing repository config | Reads .conductor/settings.toml, settings.local.toml, conductor.json and .worktreeinclude as they are. | | Scripts that read CONDUCTOR\_\* | Sets the CONDUCTOR\_\* variables Conductor gave local scripts, with the same meanings, alongside their MICROMANAGER\_\* twins. | | No starting over | One-click, read-only import of repositories, workspaces, chats and agent paths on the Mac; worktrees stay put. | | Review before merge | Diff view with line comments sent back to the agent; agents can comment too. | | Shipping | Create, check and merge pull requests through the GitHub CLI. | | A way back from bad turns | Checkpoints before every message, with Revert to here. | | Privacy | No account, no telemetry, nothing stored on a server. | | A Windows or Linux machine | The same app runs on Windows and Linux, which Conductor never supported. | [See the detailed comparison with Conductor](https://micromanager.win/compare/conductor) ## How long does switching take? As long as the import: install microManager on your Mac, choose **Import from Conductor**, check the preview and import. Your workspaces open in their existing worktrees with their chats, and in a Claude Code or Codex chat your next message continues the same agent session. The [migration guide](https://micromanager.win/migrate-from-conductor) lists exactly what carries over and what does not; cloud workspaces need Conductor’s “Transfer to local” first, as [exporting your workspaces](https://micromanager.win/export-conductor-workspaces) explains. ## Which agents can I bring? [Claude Code](https://micromanager.win/agents/claude-code), [Codex](https://micromanager.win/agents/codex), [Grok Build](https://micromanager.win/agents/grok), [OpenCode](https://micromanager.win/agents/opencode) and [Gemini CLI](https://micromanager.win/agents/gemini-cli) are built in, and any agent that speaks the [Agent Client Protocol](https://micromanager.win/features/acp-agents) can be added. microManager runs the CLIs you already have installed and signed in. ## Is there a Conductor alternative for Windows or Linux? Yes. Conductor’s desktop app is Mac-only, and microManager runs on Windows and Linux as well, with the same workspaces, agents and settings files. See [Conductor for Windows](https://micromanager.win/conductor-for-windows) and [Conductor for Linux](https://micromanager.win/conductor-for-linux). ## What are the other options? Several tools run coding agents in parallel, from terminal multiplexers to desktop apps and the agent vendors’ own apps. [The alternatives roundup](https://micromanager.win/alternatives) lists them with platforms, supported agents, licenses and prices, each verified on the tool’s own site and dated. ## Questions about replacing Conductor ### Why do I need a Conductor alternative? Meta acquired Conductor on 4 November 2026 and stopped new Mac app downloads. Conductor Cloud is read-only from 13 November 2026 and shut down from 20 November 2026, when Conductor starts deleting the data stored there, and the local Mac app stops getting new models and fixes in early 2027, when its user data is deleted. ### Do I have to set up my repositories again? No. The import brings repositories and workspaces over, and microManager reads the settings files already in your repositories, so setup and run scripts work without changes. ### Does microManager have a cloud version like Conductor Cloud? No. microManager runs everything on your own computer. It has no cloud workspaces, mobile app or team features. --- title: "Migrate from Conductor to microManager, step by step" description: "Move from Conductor to microManager: import repos, workspaces and chats in one click, keep settings.toml and CONDUCTOR_* scripts, see what stays behind." url: "https://micromanager.win/migrate-from-conductor" updated: "2026-11-08" --- # Move from Conductor to microManager Install microManager, choose Import from Conductor, review the preview and import: your repositories, workspaces and chats come over in one step, every worktree stays where it is, and your .conductor/settings.toml scripts keep working with the CONDUCTOR\_\* variables they read. Only cloud workspaces need moving to your Mac first. Updated 8 November 2026 ## How do I move over? 1. ### Move cloud workspaces to your Mac If you used Conductor Cloud, right-click each cloud workspace in Conductor and choose Transfer to local, or export your data from Settings → Account → Export data. Transfers are safest before Conductor Cloud is read-only, from 13 November 2026. Local workspaces need nothing. 2. ### Install microManager On the Mac where you use Conductor, download the DMG from https\://micromanager.win/download/mac (Apple Silicon) or https\://micromanager.win/download/mac-intel and drag microManager to Applications, or run curl -fsSL https\://micromanager.win/install.sh | sh in Terminal. `curl -fsSL https://micromanager.win/install.sh | sh` 3. ### Check that your agents are found On first launch microManager lists the agent CLIs it found on your login shell PATH (Claude Code, Codex, Grok Build, OpenCode, Gemini CLI) and their versions. Executable paths you set in Conductor are imported in the next step. 4. ### Choose Import from Conductor and review the preview microManager reads Conductor’s local database read-only and shows your repositories (and whether each folder still exists), how many active and archived workspaces and chats it found, and whether it found your Conductor settings. Choose which repositories to bring, and whether to include archived workspaces (off by default) and chat history (on by default). 5. ### Import Repositories, workspaces, chat tabs and settings are added to microManager. Each workspace keeps its existing worktree folder and branch, and nothing on disk is moved; the only file written is a .context/ line in a repository’s git exclude file, if it lacks one. Conductor’s data is never written to, and running the import again only adds what is missing. 6. ### Open a workspace and keep going Open any imported workspace: its chats show their earlier messages, and your next message continues the same Claude Code or Codex session. Setup and run scripts from .conductor/settings.toml work as before, with CONDUCTOR\_\* variables set. ![Settings, Import from Conductor: 10 active and 22 archived workspaces, 16 chats and settings found, four repositories to choose from, and switches for chat history and archived workspaces.](https://micromanager.win/screenshots/conductor-import-preview.png) The Conductor import preview, here with demo Conductor data. Conductor’s own data is only read. ## What carries over, and what does not? | From Conductor | Carries over | Notes | | ------------------------------------------------------------- | --------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Repositories | Yes | Name, folder, default branch and remote. Repositories hidden in Conductor, or whose folder is gone, are skipped. | | Active workspaces | Yes | Same folder and branch, with name, target branch, status, pin and notes. Skipped if the folder no longer exists. | | Archived workspaces | If you choose | Switch on “Include archived workspaces” in the preview; they come over archived, ready to restore. | | Chat tabs | Yes | Visible tabs, with agent, model, effort level and permission mode. Chats with Cursor, which microManager does not run, are skipped. | | Chat history | Claude Code and Codex | Read from the agents’ own transcript files, newest 2,000 events per chat; the next message resumes the same session. | | Agent executable paths | Yes | Claude Code, Codex and OpenCode paths from \~/.conductor/settings.toml become the commands of those profiles. | | Prompts, branch prefix, default model | Yes | From \~/.conductor/settings.toml: the general and action prompts, branch prefix, archive on merge, default model and effort levels, and whether tool approvals were off. Only settings still at microManager’s defaults are filled in. | | .conductor/settings.toml, settings.local.toml, conductor.json | Read in place | Same keys, read directly from your repository. Nothing to convert. | | .worktreeinclude and file_include_globs | Read in place | Ignored files are copied into new workspaces the same way. | | CONDUCTOR\_\* environment variables | Yes | The ones Conductor gave local scripts, with the same meanings, alongside the MICROMANAGER\_\* names for scripts, terminals and agents. Each workspace gets a new block of ports; the cloud and API variables are not set. | | MCP tool names | Same names | AskUserQuestion, GetWorkspaceDiff, DiffComment, GetDiffComments and GetTerminalOutput. | | Conductor Cloud workspaces | No | Transfer them to local in Conductor first; then they import like any local workspace. | | Account, billing, iPhone app | No | microManager has no account, no paid tier and no mobile app. | ## Which environment variables match? With Conductor-compatible environment on (the default), both names are set to the same value for every script, terminal and agent. | Conductor name | microManager name | Value | | -------------------------- | ----------------------------- | -------------------------------------------------------------------------------------------------------------------------------------- | | `CONDUCTOR_WORKSPACE_NAME` | `MICROMANAGER_WORKSPACE_NAME` | The workspace’s folder name, a city such as lisbon. It stays the same when you rename the workspace or its branch. | | `CONDUCTOR_WORKSPACE_ID` | `MICROMANAGER_WORKSPACE_ID` | The workspace’s unique id. | | `CONDUCTOR_WORKSPACE_PATH` | `MICROMANAGER_WORKSPACE_PATH` | Absolute path of the workspace’s git worktree. | | `CONDUCTOR_ROOT_PATH` | `MICROMANAGER_ROOT_PATH` | Absolute path of the repository’s main checkout, where files to copy come from. | | `CONDUCTOR_DEFAULT_BRANCH` | `MICROMANAGER_DEFAULT_BRANCH` | The repository’s default branch. | | `CONDUCTOR_PORT` | `MICROMANAGER_PORT` | The first of ten ports reserved for this workspace (PORT to PORT+9). Each workspace gets the lowest free multiple of 10 from 51000 up. | | `CONDUCTOR_IS_LOCAL` | `MICROMANAGER_IS_LOCAL` | Always 1: workspaces run on your own computer. | ## Should I rename my settings files? You do not have to. `.conductor/settings.toml` keeps working. If you want microManager-only settings, add `.micromanager/settings.toml`: it takes precedence over the Conductor file, key by key, so you can move one script at a time. The [repository settings](https://micromanager.win/docs/repository-settings) page lists the full order. ## How do I save cloud workspaces? Transfer them to local, or download them with Export data. Transfers are safest before Conductor Cloud is read-only, from 13 November 2026, and Conductor deletes cloud data from 20 November 2026. [How to export your Conductor workspaces](https://micromanager.win/export-conductor-workspaces) covers both, plus saving local worktrees with git. ## What if I am moving to Windows or Linux? The import runs on the Mac, where Conductor’s data is. To move to a PC instead, push the branches you want to keep, clone the repositories there and add them to microManager; your `.conductor/settings.toml` scripts and `CONDUCTOR_*` variables work there too. See [Conductor for Windows](https://micromanager.win/conductor-for-windows) and [Conductor for Linux](https://micromanager.win/conductor-for-linux). ## Can I keep Conductor installed while I try microManager? Yes, and it is a good idea while the local Conductor app still works. Both apps can show the same workspaces because the import does not move anything. Just avoid running agents in the same worktree from both apps at the same time. ## Questions about moving ### Does importing change anything in Conductor? No. microManager opens Conductor’s database read-only and never writes to it or to Conductor’s settings. Conductor keeps working exactly as before. ### Are my worktrees moved or copied? No. Each imported workspace points at the worktree folder Conductor created, on the same branch, with uncommitted work untouched. ### Do I need to rewrite my setup scripts? No. microManager reads the same settings files and sets the CONDUCTOR\_\* environment variables Conductor gave local scripts, such as CONDUCTOR_ROOT_PATH and CONDUCTOR_WORKSPACE_NAME, with the same meanings, next to its own MICROMANAGER\_\* names. Only CONDUCTOR_PORT changes, because each workspace gets a new block of ports. ### Can I run the import more than once? Yes. Records that were already imported are recognized and skipped, so a second run only adds new repositories, workspaces or chats. ### What about chats with Grok or OpenCode? They are imported as chats with the matching agent profile (the Grok profile for Grok models, OpenCode for the rest) so you can continue in the same workspace. Earlier messages are imported for Claude Code and Codex chats, which keep their own transcript files. --- title: "Export Conductor workspaces before the shutdown" description: "How to export Conductor workspaces before the shutdown: Transfer to local for cloud workspaces, Export data, and saving every local worktree in git." url: "https://micromanager.win/export-conductor-workspaces" updated: "2026-11-08" --- # How to export your Conductor workspaces Move each Conductor Cloud workspace to your Mac with “Transfer to local”, safest before the cloud is read-only from 13 November 2026, and download everything with Settings → Account → Export data; Conductor deletes cloud data from 20 November 2026. Local workspaces are git worktrees, so committing and pushing their branches saves the code itself. Updated 8 November 2026 ## How do I export everything, step by step? 1. ### Move each cloud workspace to your Mac In Conductor, right-click a cloud workspace and choose Transfer to local. It comes to your Mac with its chats and becomes a local workspace. Conductor has not said whether this works once Conductor Cloud is read-only, from 13 November 2026, so that is the safe deadline. 2. ### Download an archive of your data In Conductor, open Settings → Account → Export data to download an archive of everything. Conductor says exports keep working while the cloud is read-only, until cloud data is deleted from 20 November 2026. 3. ### Find every local workspace Conductor workspaces are git worktrees of your repositories. In each repository, git worktree list prints every worktree folder and its branch, and git status --short in a folder shows what is not committed yet. 4. ### Commit and push what you want to keep Commit the work in each workspace you care about and push its branch. Once a branch is on your remote, it no longer depends on Conductor or on that folder. 5. ### Keep working in microManager (optional) On a Mac, microManager’s Import from Conductor reads Conductor’s local data read-only and opens the same worktrees with their chats, including the workspaces you transferred from the cloud. ## What are the deadlines? | Date | What changes | What to have done | | -------------------------- | ---------------------------------------------------------------------------- | ------------------------------------------------------------------------------------ | | Friday 13 November 2026 | Conductor Cloud read-only: chats can be read and exported, not continued. | Cloud workspaces transferred to local. | | Friday 20 November 2026 | Conductor Cloud shut down; deletion of cloud data begins. | Everything you need from the cloud exported. | | Early 2027 (no exact date) | The local Mac app stops getting updates and Conductor deletes its user data. | Branches you care about committed and pushed; your daily work moved to another tool. | The [shutdown timeline](https://micromanager.win/conductor-shutdown) has each date with its sources. ## How do I find and save my local workspaces? Conductor’s documentation says it creates each workspace as a git worktree on its own branch, so your repository already knows about every one of them. In each repository: ```sh # In each repository: every worktree, its folder and its branch git worktree list # In a workspace folder: anything not yet committed? git status --short ``` For each workspace you want to keep, commit and push its branch: ```sh # In the workspace folder: commit the work and push the branch git add -A git commit -m "Work in progress from Conductor" git push -u origin HEAD ``` Pushed branches are safe whatever happens to Conductor. Notes in a workspace’s `.context` folder are not committed, so copy any you want to keep. ## What about my chat history? Cloud chats come to your Mac with Transfer to local, and Export data downloads everything. For local workspaces, the agents keep their own transcripts: Claude Code under `~/.claude/projects` and Codex under `~/.codex/sessions`. microManager’s import reads those, so imported chats show their earlier messages and continue the same agent sessions. ## Where should the workspaces go next? If you want to keep working the same way, [microManager](https://micromanager.win/) imports your local Conductor repositories, workspaces and chats in one click and leaves every worktree where it is; the [migration guide](https://micromanager.win/migrate-from-conductor) covers each step. To compare other tools, see [the alternatives](https://micromanager.win/alternatives). ## Sources 1. [Conductor is joining Meta](https://www.conductor.build/announcement), Conductor. Read 8 November 2026. 2. [Meta acquires AI coding startup Conductor, whose app runs several coding agents at once](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), SiliconANGLE, Duncan Riley, 4 November 2026. Read 8 November 2026. 3. [Conductor documentation: workspaces and branches](https://www.conductor.build/docs/concepts/workspaces-and-branches), Conductor. Read 8 November 2026. ## Questions about exporting ### How do I export a Conductor Cloud workspace? Right-click it in Conductor and choose Transfer to local, which moves it and its chats to your Mac, or use Settings → Account → Export data to download an archive of everything. Conductor Cloud is read-only from 13 November 2026 and shut down from 20 November 2026, when Conductor starts deleting the data stored there. ### Can I export once Conductor Cloud is read-only? Yes, until cloud data is deleted from 20 November 2026. Conductor says read-only cloud workspaces can be read and exported, but not sent new messages. It has not said whether Transfer to local works once the cloud is read-only, from 13 November 2026. ### Do local Conductor workspaces need exporting? Not to keep your code. A local workspace is a git worktree folder on your Mac, so its commits live in your repository. Commit and push anything you want to keep, because uncommitted changes exist only in that folder. ### Can microManager import the Export data archive? No. microManager imports from the Conductor data on your Mac, read-only, not from the archive. Transfer cloud workspaces to local first; they then import like any other local workspace. ### What happens to my local Conductor data in early 2027? Conductor says the Mac app keeps getting new models and critical fixes until early 2027, and that it will delete Conductor user data when the app shuts down then. It has not given an exact date. --- title: "Conductor shutdown timeline: dates and what to do" description: "Conductor shutdown dates: Meta deal 4 Nov 2026, Cloud read-only 13 Nov, Cloud closed 20 Nov, Mac app winds down early 2027. What to do, with sources." url: "https://micromanager.win/conductor-shutdown" updated: "2026-11-08" --- # What happens to Conductor after the Meta acquisition Meta acquired Conductor on 4 November 2026; Conductor Cloud is read-only from 13 November 2026 and shut down from 20 November 2026, when Conductor starts deleting the data stored there. The local Mac app keeps working, with new models and critical fixes, until early 2027, then winds down and its user data is deleted. Updated 8 November 2026 ## What are the dates? From Conductor’s own announcement and reporting by TechCrunch and SiliconANGLE, read on 8 November 2026. Each date is marked as it passes. 1. 4 November 2026 Wednesday ### Meta acquires Conductor; sign-ups and Mac downloads stop Conductor announced it is joining Meta, and TechCrunch and SiliconANGLE reported the acquisition. The same day, new users could no longer sign up for Conductor or download its Mac app; its iPhone app stayed on the App Store. Free users can no longer upgrade, and Pro subscribers keep their paid features at no charge. Sources: [TechCrunch](https://techcrunch.com/2026/11/04/meta-acquires-ai-coding-startup-conductor/), [SiliconANGLE](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), [Conductor](https://www.conductor.build/announcement) 2. 13 November 2026 Friday ### Conductor Cloud goes read-only Cloud workspaces stop accepting new messages. Old chats can still be read and exported, and active subscribers get refunds. Sources: [SiliconANGLE](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), [Conductor](https://www.conductor.build/announcement) 3. 20 November 2026 Friday ### Conductor Cloud shuts down The cloud service closes and Conductor begins deleting the data users stored there. Sources: [TechCrunch](https://techcrunch.com/2026/11/04/meta-acquires-ai-coding-startup-conductor/), [SiliconANGLE](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), [Conductor](https://www.conductor.build/announcement) 4. Early 2027 Exact date not announced ### The local Mac app winds down Local copies of the Mac app keep working and keep getting new models and critical fixes until early 2027. After that, updates stop and Conductor deletes the app’s user data. Sources: [SiliconANGLE](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), [Conductor](https://www.conductor.build/announcement) ## What happened on 4 November 2026? Conductor, the Mac app for running several coding agents at once, announced that it is joining Meta. Its maker is Melty, Inc., as Conductor’s [privacy policy](https://www.conductor.build/privacy) and [App Store listing](https://apps.apple.com/us/app/conductor-build/id6791228564) name it (its [terms of service](https://www.conductor.build/terms) say Melty Labs, Inc.). [TechCrunch](https://techcrunch.com/2026/11/04/meta-acquires-ai-coding-startup-conductor/) and [SiliconANGLE](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/) reported the acquisition the same day; the terms were not disclosed. Conductor stopped new sign-ups and downloads of its Mac app immediately. Free users can no longer upgrade, and Pro subscribers keep their paid features at no charge. ## What happens to Conductor Cloud? From Friday 13 November 2026, cloud workspaces are read-only: old chats can be read and data exported, but no new messages sent, and active subscribers get refunds. From Friday 20 November 2026, Conductor Cloud is shut down and Conductor deletes the data users stored there. Conductor’s announcement gives two ways out: right-click a cloud workspace and choose **Transfer to local** to move it and its chats to your Mac, or use **Settings → Account → Export data** to download everything. Conductor has not said whether transfers work once the cloud is read-only, so the safe deadline for them is 13 November 2026. [How to export your Conductor workspaces](https://micromanager.win/export-conductor-workspaces) walks through it. ## What happens to the local Conductor Mac app? Copies already installed keep working, and keep getting new models and critical bug fixes, until early 2027. Then the app winds down: updates stop and Conductor deletes the app’s user data. Conductor has not given an exact date. Because Mac app downloads stopped on 4 November 2026, you cannot install Conductor on a new or reset Mac. ## What should I do, and by when? | When | What to do | | ----------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------ | | By 13 November 2026, when Cloud is read-only | Move cloud workspaces to your Mac with “Transfer to local”, or export them. Read-only cloud chats can be read and exported, not continued. | | By 20 November 2026, when cloud data is deleted | Export anything left in Conductor Cloud (Settings → Account → Export data). | | While Conductor still runs | Install a replacement and import your local workspaces, so you can compare the two side by side. | | By early 2027 | Make sure every branch you care about is committed and pushed. Your worktrees are ordinary git, so the work itself does not depend on Conductor. | ## Are my branches and worktrees safe? Conductor’s workspaces are git worktrees ([its documentation](https://www.conductor.build/docs/concepts/workspaces-and-branches) says so), so your branches and commits live in your own repositories, not in Conductor. Uncommitted changes exist only in the worktree folders, so commit what you want to keep before you archive or delete anything. ## Where can I move? [microManager](https://micromanager.win/) is a free app for macOS, Windows and Linux built to take over from Conductor’s local app: it reads `.conductor/settings.toml` and `conductor.json`, sets the `CONDUCTOR_*` environment variables your scripts expect, and on a Mac imports your repositories, workspaces and chats in one click without moving anything. The [migration guide](https://micromanager.win/migrate-from-conductor) walks through it, and [the alternatives page](https://micromanager.win/alternatives) compares other tools, each checked against its own site. ## Sources 1. [Meta acquires AI coding startup Conductor](https://techcrunch.com/2026/11/04/meta-acquires-ai-coding-startup-conductor/), TechCrunch, Lucas Ropek, 4 November 2026. Read 8 November 2026. 2. [Meta acquires AI coding startup Conductor, whose app runs several coding agents at once](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), SiliconANGLE, Duncan Riley, 4 November 2026. Read 8 November 2026. 3. [Conductor is joining Meta](https://www.conductor.build/announcement), Conductor. Read 8 November 2026. 4. [Conductor documentation: workspaces and branches](https://www.conductor.build/docs/concepts/workspaces-and-branches), Conductor. Read 8 November 2026. 5. [Conductor privacy policy (last updated 3 November 2026)](https://www.conductor.build/privacy), Conductor, 3 November 2026. Read 8 November 2026. 6. [Conductor.build on the App Store (iPhone)](https://apps.apple.com/us/app/conductor-build/id6791228564), Apple App Store. Read 8 November 2026. 7. [Conductor terms of service (last modified 3 November 2026)](https://www.conductor.build/terms), Conductor, 3 November 2026. Read 8 November 2026. ## Questions about the shutdown ### When does Conductor shut down? Conductor Cloud is read-only from 13 November 2026 and shut down from 20 November 2026, when Conductor starts deleting the data stored there. The local Mac app keeps working until early 2027, then winds down. No exact date for the Mac app has been announced. ### Will the Conductor Mac app stop working? According to Conductor and SiliconANGLE, local copies keep working and keep getting new models and critical bug fixes until early 2027. After that, updates stop and Conductor deletes the app’s user data. ### Can I still download Conductor? Not the Mac app. New sign-ups and Mac app downloads stopped on 4 November 2026, so you cannot install it on a new Mac. Its iPhone app is still on the App Store, for people who already use Conductor. ### Will Conductor subscribers get a refund? Conductor says active subscribers will get a refund. Pro subscribers keep their Pro features at no charge. ### What will Conductor’s team build at Meta? Conductor says it will have something new early next year, and SiliconANGLE describes it as a new interface that has not yet been described. No product details have been published. --- title: "microManager vs Conductor: feature comparison" description: "microManager vs Conductor, side by side: status after the Meta deal, price, agents, worktrees, settings files, environment variables, review and PRs." url: "https://micromanager.win/compare/conductor" updated: "2026-11-08" --- # microManager vs Conductor Both run coding agents in parallel, each in its own git worktree, with review and pull requests. The difference now is availability: Conductor stopped new Mac app downloads on 4 November 2026 and winds down in early 2027, while microManager is free, local-only, reads Conductor’s settings files and imports its workspaces. Updated 8 November 2026 ## How do they compare? Conductor facts come from Conductor’s site, documentation, announcement and App Store listing and from SiliconANGLE, all read on 8 November 2026; they describe Conductor’s local Mac app unless a row says otherwise. | Topic | microManager | Conductor | | ------------------------------ | --------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------- | | Status (as of 8 November 2026) | Available to download. | No new sign-ups or downloads since 4 November 2026; the Mac app gets new models and critical fixes until early 2027. | | Price | Free. No account and no paid tier. | Had a Pro plan; existing Pro users keep Pro features at no charge, and free users can no longer upgrade. | | Where it runs | On your own computer: macOS (Apple Silicon, and Intel when a release has an Intel build), Windows x64 or Linux x64. | A Mac app, plus Conductor Cloud (closing on 20 November 2026) and an iPhone app for controlling cloud agents. | | Built-in agents | Claude Code, Codex, Grok Build, OpenCode, Gemini CLI, and any ACP agent you add. | Claude Code, Codex, Cursor and OpenCode. | | Isolation | One git worktree and branch per workspace. | One git worktree and branch per workspace. | | Repository settings | Reads .conductor/settings.toml, settings.local.toml, conductor.json and .worktreeinclude, plus its own .micromanager files. | Its own .conductor settings files. | | Environment variables | MICROMANAGER\_\* and the same values as CONDUCTOR\_\* (such as CONDUCTOR_PORT). | CONDUCTOR\_\* (such as CONDUCTOR_PORT). | | Reviewing and shipping | Diff view with comments sent back to the agent, agent review comments, pull request creation, checks and merge. | A diff viewer, checks, pull requests and merge. | | Your data later | Stays in a local database on your computer; no server copy. | Cloud data deletion begins 20 November 2026; the Mac app’s user data is deleted when it winds down in early 2027. | ## What else does microManager include? - [Checkpoints](https://micromanager.win/features/checkpoints) before every message, with Revert to here. - [Native ACP support](https://micromanager.win/features/acp-agents), so Grok Build runs with `grok agent stdio` and no adapter. - [MCP tools](https://micromanager.win/docs/mcp-tools) named like Conductor’s: AskUserQuestion, GetWorkspaceDiff, DiffComment, GetDiffComments and GetTerminalOutput. - A one-click, read-only [import from Conductor](https://micromanager.win/migrate-from-conductor). ## Which should I use? If Conductor’s local app works for you today, you can keep using it until early 2027. If you want something that will keep getting updates, need to set up a new Mac, or work on Windows or Linux, microManager takes over your existing workspaces without changing them, and on a Mac you can run both side by side while you decide. If you relied on Conductor Cloud or its iPhone app, look at [the alternatives](https://micromanager.win/alternatives), since microManager has neither. ## Sources 1. [Conductor is joining Meta](https://www.conductor.build/announcement), Conductor. Read 8 November 2026. 2. [Conductor home page](https://www.conductor.build/), Conductor. Read 8 November 2026. 3. [Conductor documentation: introduction](https://www.conductor.build/docs/), Conductor. Read 8 November 2026. 4. [Conductor documentation: workspaces and branches](https://www.conductor.build/docs/concepts/workspaces-and-branches), Conductor. Read 8 November 2026. 5. [Conductor.build on the App Store (iPhone)](https://apps.apple.com/us/app/conductor-build/id6791228564), Apple App Store. Read 8 November 2026. 6. [Meta acquires AI coding startup Conductor, whose app runs several coding agents at once](https://siliconangle.com/2026/11/04/meta-acquires-ai-coding-startup-conductor-whose-app-runs-several-coding-agents-at-once/), SiliconANGLE, Duncan Riley, 4 November 2026. Read 8 November 2026. 7. [Meta acquires AI coding startup Conductor](https://techcrunch.com/2026/11/04/meta-acquires-ai-coding-startup-conductor/), TechCrunch, Lucas Ropek, 4 November 2026. Read 8 November 2026. ## Questions about the comparison ### Is microManager a fork or copy of Conductor? No. microManager is an independent app with its own code and design. It reads Conductor’s settings file formats and environment variable names so that existing repositories and scripts keep working. ### Does microManager support Cursor like Conductor did? Not as a built-in profile. The built-in agents are Claude Code, Codex, Grok Build, OpenCode and Gemini CLI, and any agent that speaks the Agent Client Protocol can be added as a custom profile. ### Is there anything Conductor had that microManager does not? Yes: Conductor Cloud, where agents ran in the cloud, an iPhone app for controlling them, and a built-in Cursor agent. microManager runs only on your own computer, on macOS, Windows or Linux, and has no cloud, mobile app or Cursor profile. --- title: "Conductor for Windows? A free alternative: microManager" description: "Conductor has no Windows app. microManager runs Claude Code, Codex and other agents in parallel git worktrees on Windows, free, with Conductor's settings." url: "https://micromanager.win/conductor-for-windows" updated: "2026-11-08" --- # A Conductor alternative for Windows Conductor is a Mac app with no Windows version, so a Windows developer who wants its way of working needs another tool. microManager is a free Windows app that runs Claude Code, Codex, Grok Build, OpenCode and Gemini CLI in parallel, each in its own git worktree, and reads Conductor’s repository settings files. Updated 8 November 2026 [Download for Windows](https://micromanager.win/download/windows) [Install steps](https://micromanager.win/install#windows) ## Is there a Conductor app for Windows? No. Conductor’s own site and documentation describe a Mac app, and offer no Windows or Linux version (we checked on 8 November 2026). Conductor also offered Conductor Cloud, shut down from 20 November 2026, and an iPhone app. Since Meta acquired Conductor on 4 November 2026, nobody new can sign up or download its Mac app; the [shutdown timeline](https://micromanager.win/conductor-shutdown) has the details. ## What does microManager do on Windows? The same things it does on a Mac, with Windows’ own shells, apps and keys. | Part | On Windows | | --------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | | Parallel agents | Many agent sessions at once, across repositories, each task in its own workspace with its own chat tabs. | | Isolation | Every workspace is a git worktree on its own branch, with ten reserved ports and your ignored files (such as .env) copied in. | | Scripts | Setup, run and archive scripts run with Git Bash when Git for Windows is installed, so the same bash scripts work, otherwise PowerShell. | | Terminals | A terminal drawer per workspace, using PowerShell 7 when installed, then Windows PowerShell, then the command prompt. | | Open in | File Explorer, Windows Terminal, VS Code, Cursor, Zed, Windsurf, or the branch on GitHub. | | Review and ship | Diff review with line comments sent back to the agent, checkpoints before every message, and pull requests through the GitHub CLI. | | Notifications | A Windows notification when an agent finishes or needs you, and a dot on the taskbar button, which flashes. | | Shortcuts | Ctrl takes the place of the Mac’s ⌘: Ctrl+N new workspace, Ctrl+T new tab, Ctrl+K command palette. | ## Which coding agents can I run on Windows? [Claude Code](https://micromanager.win/agents/claude-code), [Codex](https://micromanager.win/agents/codex), [Grok Build](https://micromanager.win/agents/grok), [OpenCode](https://micromanager.win/agents/opencode), [Gemini CLI](https://micromanager.win/agents/gemini-cli) and any other [ACP agent](https://micromanager.win/features/acp-agents), as long as the agent’s own CLI is installed and signed in on your PC. microManager starts the CLIs you already have; it finds npm-installed commands such as `claude.cmd` on your `PATH` by itself. With Git for Windows installed, Claude Code runs its shell commands in Git Bash; without it, in PowerShell. ## Will scripts written for Conductor work on Windows? microManager reads `.conductor/settings.toml`, `conductor.json` and `.worktreeinclude` on Windows too, and sets the same `CONDUCTOR_*` variables. With Git for Windows installed, scripts run in Git Bash, so most bash scripts written on a Mac work as they are. Commands that only exist on macOS, such as `open`, `pbcopy` or `brew`, need a Windows equivalent; put a Windows-only version in `.micromanager/settings.local.toml` on that PC, which takes precedence (see [repository settings](https://micromanager.win/docs/repository-settings)). ## How do I install microManager on Windows? Paste `irm https://micromanager.win/install.ps1 | iex` into PowerShell: it downloads the installer, checks its SHA-512 against the release manifest and installs microManager for your user without administrator rights. Or download the installer and run it yourself; it is not code-signed yet, so SmartScreen may warn you: check its SHA-512 first. The [download page](https://micromanager.win/install#windows) has the steps, requirements and the PowerShell lines to check the file. ## I used Conductor on a Mac. How do I move my work to Windows? Conductor’s workspaces are ordinary git branches, so push the ones you want to keep from the Mac, clone the repository on your PC and add it to microManager; then **Ctrl+Shift+N** starts a workspace on each branch, which it checks out as it is. Chats stay on the Mac: Import from Conductor reads Conductor’s local data, which exists only on a Mac. If you will keep a Mac too, [import there](https://micromanager.win/migrate-from-conductor) to keep your chats. ## Sources 1. [Conductor home page](https://www.conductor.build/), Conductor. Read 8 November 2026. 2. [Conductor documentation: introduction](https://www.conductor.build/docs/), Conductor. Read 8 November 2026. 3. [Conductor.build on the App Store (iPhone)](https://apps.apple.com/us/app/conductor-build/id6791228564), Apple App Store. Read 8 November 2026. 4. [Conductor is joining Meta](https://www.conductor.build/announcement), Conductor. Read 8 November 2026. ## Questions about Windows ### Is there a Conductor app for Windows? No. Conductor describes itself as an app for running coding agents on a Mac, and its site offers no Windows version. Since 4 November 2026 it has also stopped taking new users on any system. ### Is microManager free on Windows? Yes. It is free on every system, with no account and no paid tier. You pay your agent providers directly. ### Does the installer need administrator rights? No. It installs for your user only. ### Why does Windows warn about the installer? The Windows installer is not code-signed yet, so Microsoft Defender SmartScreen may say it comes from an unknown publisher. Check the file’s SHA-512 against latest.yml as the download page shows, then choose More info and Run anyway. ### Does it run on Windows on ARM? Not yet. The Windows build is for x64 PCs. ### Can I import my Conductor workspaces on Windows? No. Conductor’s desktop app only ever ran on Macs, so that is where its data is. Import from Conductor is available in the Mac version; on Windows, clone your repositories and add them. --- title: "Conductor for Linux? A free alternative: microManager" description: "Conductor has no Linux app. microManager runs Claude Code, Codex, Grok Build and more in parallel git worktrees on Linux, as a free AppImage or .deb." url: "https://micromanager.win/conductor-for-linux" updated: "2026-11-08" --- # A Conductor alternative for Linux Conductor is a Mac app with no Linux version, so on Linux you need another tool for its way of working. microManager is a free Linux app, as an AppImage or a .deb, that runs Claude Code, Codex, Grok Build, OpenCode and Gemini CLI in parallel, each in its own git worktree, and reads Conductor’s repository settings files. Updated 8 November 2026 [Download the AppImage](https://micromanager.win/download/linux) [Download the .deb](https://micromanager.win/download/linux-deb) [Install steps](https://micromanager.win/install#linux) ## Is there a Conductor app for Linux? No. Conductor’s own site and documentation describe a Mac app and offer no Linux or Windows version (we checked on 8 November 2026). Since Meta acquired Conductor on 4 November 2026, nobody new can sign up or download its Mac app, and Conductor Cloud is shut down from 20 November 2026; the [shutdown timeline](https://micromanager.win/conductor-shutdown) has the details. ## What does microManager do on Linux? The same things it does on a Mac, with your Linux shell, desktop and apps. | Part | On Linux | | --------------- | ------------------------------------------------------------------------------------------------------------------------------------- | | Packages | x64, as an AppImage that runs on most distributions and a .deb for Debian and Ubuntu. | | Isolation | Every workspace is a git worktree on its own branch, with ten reserved ports and your ignored files (such as .env) copied in. | | Scripts | Setup, run and archive scripts run with your login shell when it is zsh or bash, otherwise bash, with your login shell’s environment. | | Terminals | A terminal drawer per workspace running your $SHELL, or bash. | | Open in | your file manager, your default terminal, VS Code, Cursor, Zed, Windsurf, or the branch on GitHub. | | Review and ship | Diff review with line comments sent back to the agent, checkpoints before every message, and pull requests through the GitHub CLI. | | Shortcuts | Ctrl takes the place of the Mac’s ⌘: Ctrl+N new workspace, Ctrl+T new tab, Ctrl+K command palette. | ## Which coding agents can I run on Linux? [Claude Code](https://micromanager.win/agents/claude-code), [Codex](https://micromanager.win/agents/codex), [Grok Build](https://micromanager.win/agents/grok), [OpenCode](https://micromanager.win/agents/opencode), [Gemini CLI](https://micromanager.win/agents/gemini-cli) and any other [ACP agent](https://micromanager.win/features/acp-agents) whose CLI is installed and signed in. microManager looks them up on your login shell’s `PATH`, the same one a new terminal gets, or uses the absolute path you set in Settings → Agents. ## Will scripts written for Conductor work on Linux? Usually, yes. microManager reads `.conductor/settings.toml`, `conductor.json` and `.worktreeinclude`, sets the same `CONDUCTOR_*` variables, and runs scripts with bash or zsh. Only commands that exist just on macOS, such as `open`, `pbcopy` or `brew`, need a Linux equivalent (`xdg-open`, `xclip`, your package manager). Put a Linux-only version in `.micromanager/settings.local.toml`, which takes precedence (see [repository settings](https://micromanager.win/docs/repository-settings)). ## How do I install microManager on Linux? Paste `curl -fsSL https://micromanager.win/install.sh | sh` into a terminal: it downloads the AppImage, checks its SHA-512, installs it as `~/.local/bin/micromanager` and adds it to your applications menu, without sudo. Or download the AppImage, make it executable with `chmod +x` and run it, or install the .deb with `sudo apt install ./FILE.deb`. The [download page](https://micromanager.win/install#linux) has the steps, requirements and how to check the file’s SHA-512. ## I used Conductor on a Mac. How do I move my work to Linux? Conductor’s workspaces are ordinary git branches, so push the ones you want to keep from the Mac, clone the repository on Linux and add it to microManager; then **Ctrl+Shift+N** starts a workspace on each branch, which it checks out as it is. Chats stay on the Mac, because Import from Conductor reads Conductor’s local data, which exists only there. If you keep a Mac as well, [import there](https://micromanager.win/migrate-from-conductor) to keep your chats. ## Sources 1. [Conductor home page](https://www.conductor.build/), Conductor. Read 8 November 2026. 2. [Conductor documentation: introduction](https://www.conductor.build/docs/), Conductor. Read 8 November 2026. 3. [Conductor is joining Meta](https://www.conductor.build/announcement), Conductor. Read 8 November 2026. ## Questions about Linux ### Is there a Conductor app for Linux? No. Conductor describes itself as an app for running coding agents on a Mac, and its site offers no Linux version. Since 4 November 2026 it has also stopped taking new users on any system. ### Which Linux distributions does microManager support? The AppImage runs on most x64 desktop distributions without installing anything, and the .deb installs on Debian, Ubuntu and their derivatives. ARM64 Linux is not supported yet. ### Does it need root? The AppImage does not: make it executable and run it. Installing the .deb goes through your package manager, which asks for sudo as usual. ### Does it update itself on Linux? Yes. The AppImage and the .deb check for a new version when they start and every six hours, like the Mac and Windows apps, and show Restart to update when it is ready. Updating the .deb asks for your password. ### Can I import my Conductor workspaces on Linux? No. Conductor’s desktop app only ever ran on Macs, so that is where its data is. Import from Conductor is available in the Mac version; on Linux, clone your repositories and add them. --- title: "Conductor alternatives: parallel agent tools (Nov 2026)" description: "Parallel coding agent tools compared from their own sites (Nov 2026): Orca, Paradev, Emdash, Agentastic, Nimbalyst, Claude Squad and more." url: "https://micromanager.win/alternatives" updated: "2026-11-08" --- # Conductor alternatives: tools for running coding agents in parallel Since Meta bought Conductor on 4 November 2026, the main choices for running several coding agents at once are desktop apps that give each agent its own git worktree (microManager, Orca, Paradev, Emdash, Agentastic, Nimbalyst), terminal tools such as Claude Squad, and the single-agent apps from Anthropic and OpenAI. Every fact below comes from each tool’s own site or repository, read on 8 November 2026. Updated 8 November 2026 ## How do they compare at a glance? microManager is ours, so it is listed first and marked; everything else is in no particular order. Tools change quickly: check each one’s site before you decide. | Tool | Platforms | Agents | Isolation | License and price | | -------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | [microManager](https://micromanager.win/) (this site) | macOS (Apple Silicon; Intel when a release has an Intel build), Windows x64, Linux x64 | Claude Code, Codex, Grok Build, OpenCode, Gemini CLI and any ACP agent | A git worktree and branch per workspace | Free, closed source, no account; imports from Conductor | | [Orca](https://github.com/stablyai/orca) | macOS (Apple Silicon and Intel), Windows, Linux; iOS and Android companion apps | Any CLI agent; its README names 34, including Claude Code, Codex, Grok, OpenCode, Cursor and GitHub Copilot | Each agent works in its own git worktree; a worktree can also get its own cloud VM | MIT, free and open source; you bring your own agent subscriptions | | [Paradev (formerly Superset)](https://paradev.com/blog/a-new-name) | macOS, with an experimental Linux build; an iPhone and iPad app | Claude Code, Codex, OpenCode, Cursor’s agent, Copilot CLI and other CLI agents | Each task runs in its own git worktree workspace | MIT since 4 November 2026 (previously the Elastic License 2.0); uses your own subscriptions and API keys and does not bill model usage; the mobile app needs Paradev Pro | | [Emdash](https://github.com/generalaction/emdash) | macOS (Apple Silicon and Intel), Windows, Linux; remote machines over SSH | Claude Code, Codex, Cursor, OpenCode, Amp, Devin, Qwen Code, Droid, GitHub Copilot and more | Every task runs in its own git worktree and branch | Apache-2.0, open source; no price listed | | [Agentastic](https://www.agentastic.dev/) | macOS 14 or later, Apple Silicon and Intel | 55 built-in agent definitions, 31 of them native or over ACP, including Claude Code, Codex, Gemini CLI and OpenCode | Each workspace is a new git worktree; Apple containers, Docker and remote runners are optional | Free with no sign-up; closed development for now | | [Nimbalyst (formerly Crystal)](https://github.com/Nimbalyst/nimbalyst) | macOS (Apple Silicon and Intel), Windows, Linux; iOS and Android companion apps | Claude Code, Codex, OpenCode and other agents | Git worktree isolation for parallel sessions | MIT, free and open source | | [Claude Squad](https://github.com/smtg-ai/claude-squad) | Installs with Homebrew or a shell script; needs tmux and gh | Claude Code, Codex, Gemini, Aider and other local terminal agents | A tmux session per agent and a git worktree per task | AGPL-3.0, free | | [T3 Code](https://github.com/pingdotgg/t3code) | macOS, Windows and Linux desktop builds; iOS and Android; a web app | Codex, Claude, Cursor, Grok Build, OpenCode, Antigravity and fx | Its README does not describe worktrees | MIT; nothing for sale; uses your existing provider plans | | [Claude Code in the Claude desktop app](https://code.claude.com/docs/en/desktop) | macOS, Windows, Linux (beta); local, cloud, SSH and WSL sessions | Claude only | Optional per-session git worktree (off by default) under .claude/worktrees; reads .worktreeinclude | Part of Claude; sign in with a Claude account or a supported provider | | [Codex in the ChatGPT desktop app](https://learn.chatgpt.com/docs/app) | macOS, Windows and Linux (the ChatGPT desktop app) | Codex only | A managed git worktree per chat under $CODEX_HOME/worktrees, with handoff to your local checkout; reads .worktreeinclude | Sign in with a ChatGPT account | | [Vibe Kanban](https://github.com/BloopAI/vibe-kanban) | Runs locally with Node, or self-hosted | Claude Code, Codex, Gemini CLI, GitHub Copilot, Amp, Cursor, OpenCode, Droid, CCR and Qwen Code | Each workspace gets a branch, a terminal and a dev server | Apache-2.0 | ## What does each tool offer? ### Orca Stably AI Desktop app - Platforms macOS (Apple Silicon and Intel), Windows, Linux; iOS and Android companion apps - Agents Any CLI agent; its README names 34, including Claude Code, Codex, Grok, OpenCode, Cursor and GitHub Copilot - Isolation Each agent works in its own git worktree; a worktree can also get its own cloud VM - Review Line notes on diffs sent back to the agent; drafts commit messages and pull requests - License, price MIT, free and open source; you bring your own agent subscriptions Also offers SSH and remote execution, a built-in browser and a coordinator agent that spawns workers. Source: [github.com/stablyai/orca](https://github.com/stablyai/orca), read 8 November 2026. ### Paradev (formerly Superset) Superset Inc. Desktop app and CLI - Platforms macOS, with an experimental Linux build; an iPhone and iPad app - Agents Claude Code, Codex, OpenCode, Cursor’s agent, Copilot CLI and other CLI agents - Isolation Each task runs in its own git worktree workspace - Review Reply to agents and review diffs from the iPhone and iPad app - License, price MIT since 4 November 2026 (previously the Elastic License 2.0); uses your own subscriptions and API keys and does not bill model usage; the mobile app needs Paradev Pro Renamed from Superset on 26 October 2026; the superset terminal command keeps working. Sources: [paradev.com/blog/a-new-name](https://paradev.com/blog/a-new-name), [paradev.com/blog/paradev-is-now-mit](https://paradev.com/blog/paradev-is-now-mit), read 8 November 2026. ### Emdash General Action (YC W26) Desktop app - Platforms macOS (Apple Silicon and Intel), Windows, Linux; remote machines over SSH - Agents Claude Code, Codex, Cursor, OpenCode, Amp, Devin, Qwen Code, Droid, GitHub Copilot and more - Isolation Every task runs in its own git worktree and branch - Review Diff review, pull requests, CI checks and merging in the app - License, price Apache-2.0, open source; no price listed Source: [github.com/generalaction/emdash](https://github.com/generalaction/emdash), read 8 November 2026. ### Agentastic Agentastic Native macOS app - Platforms macOS 14 or later, Apple Silicon and Intel - Agents 55 built-in agent definitions, 31 of them native or over ACP, including Claude Code, Codex, Gemini CLI and OpenCode - Isolation Each workspace is a new git worktree; Apple containers, Docker and remote runners are optional - Review Native unified and side-by-side diff viewer, AI review, pull requests and merging - License, price Free with no sign-up; closed development for now Source: [agentastic.dev](https://www.agentastic.dev/), read 8 November 2026. ### Nimbalyst (formerly Crystal) Nimbalyst Desktop app - Platforms macOS (Apple Silicon and Intel), Windows, Linux; iOS and Android companion apps - Agents Claude Code, Codex, OpenCode and other agents - Isolation Git worktree isolation for parallel sessions - Review A visual workspace for agent sessions, with diffs, tasks, docs, diagrams and mockups - License, price MIT, free and open source Crystal, the earlier app from the same team, was deprecated in February 2026 in favor of Nimbalyst. Sources: [github.com/Nimbalyst/nimbalyst](https://github.com/Nimbalyst/nimbalyst), [github.com/stravu/crystal](https://github.com/stravu/crystal), read 8 November 2026. ### Claude Squad smtg-ai Terminal app (tmux) - Platforms Installs with Homebrew or a shell script; needs tmux and gh - Agents Claude Code, Codex, Gemini, Aider and other local terminal agents - Isolation A tmux session per agent and a git worktree per task - Review Terminal-based: review and push each branch yourself - License, price AGPL-3.0, free Source: [github.com/smtg-ai/claude-squad](https://github.com/smtg-ai/claude-squad), read 8 November 2026. ### T3 Code Ping (pingdotgg) Desktop (Electron), web, mobile and CLI - Platforms macOS, Windows and Linux desktop builds; iOS and Android; a web app - Agents Codex, Claude, Cursor, Grok Build, OpenCode, Antigravity and fx - Isolation Its README does not describe worktrees - Review Its README links a source control integrations page without details - License, price MIT; nothing for sale; uses your existing provider plans The README says “We are very very early in this project. Expect bugs.” Source: [github.com/pingdotgg/t3code](https://github.com/pingdotgg/t3code), read 8 November 2026. ### Claude Code in the Claude desktop app Anthropic The Code tab of the Claude desktop app - Platforms macOS, Windows, Linux (beta); local, cloud, SSH and WSL sessions - Agents Claude only - Isolation Optional per-session git worktree (off by default) under .claude/worktrees; reads .worktreeinclude - Review Diff view with line comments Claude acts on; pull requests with CI status, auto-fix and merge when ready - License, price Part of Claude; sign in with a Claude account or a supported provider Source: [code.claude.com/docs/en/desktop](https://code.claude.com/docs/en/desktop), read 8 November 2026. ### Codex in the ChatGPT desktop app OpenAI The Codex mode of the ChatGPT desktop app - Platforms macOS, Windows and Linux (the ChatGPT desktop app) - Agents Codex only - Isolation A managed git worktree per chat under $CODEX_HOME/worktrees, with handoff to your local checkout; reads .worktreeinclude - Review Its docs have a code review section - License, price Sign in with a ChatGPT account Sources: [learn.chatgpt.com/docs/app](https://learn.chatgpt.com/docs/app), [learn.chatgpt.com/docs/environments/git-worktrees](https://learn.chatgpt.com/docs/environments/git-worktrees), read 8 November 2026. ### Vibe Kanban Bloop AI until April 2026; now community maintained Kanban board with agent workspaces (runs with npx) - Platforms Runs locally with Node, or self-hosted - Agents Claude Code, Codex, Gemini CLI, GitHub Copilot, Amp, Cursor, OpenCode, Droid, CCR and Qwen Code - Isolation Each workspace gets a branch, a terminal and a dev server - Review Inline diff comments sent to the agent; pull requests with generated descriptions - License, price Apache-2.0 Bloop, the company behind it, shut down on 10 April 2026; the project goes on as community-maintained open source, and its README says “Vibe Kanban is sunsetting”. Sources: [github.com/BloopAI/vibe-kanban](https://github.com/BloopAI/vibe-kanban), [vibekanban.com/blog/shutdown](https://www.vibekanban.com/blog/shutdown), read 8 November 2026. ## How do I choose? - **Your platform.** microManager, Orca, Emdash, Nimbalyst, T3 Code and the ChatGPT app that runs Codex work on macOS, Windows and Linux; Paradev is for macOS with an experimental Linux build, and Agentastic’s app is Mac only. - **Your agents.** The Claude and ChatGPT desktop apps run only their own agent. Multi-agent tools differ in which CLIs they support out of the box and whether they speak ACP. - **Local or remote.** Some tools add SSH, containers or cloud machines; others, microManager included, run everything on your own computer. - **How much you need to move.** If you have many Conductor workspaces, a tool that imports them and reads `.conductor/settings.toml` saves you rebuilding each repository’s setup. The [Conductor alternative](https://micromanager.win/conductor-alternative) page shows how microManager does that. - **Whether it will be around.** Bloop, the company behind Vibe Kanban, shut down in April 2026 and left the project to its community, Crystal was replaced by Nimbalyst, and T3 Code calls itself very early. Check each project’s recent activity. ## What about plain git worktrees? Running `git worktree add ../my-app-task -b task` and starting an agent CLI in the new folder works with no app at all. What you give up is what the apps automate: copying `.env` files, running setup, assigning ports, watching several chats at once, reviewing diffs and cleaning up worktrees afterwards. ## Questions about the alternatives ### What can I use instead of Conductor? Desktop apps that run several coding agents in git worktrees include microManager, Orca, Paradev (formerly Superset), Emdash, Agentastic and Nimbalyst. Claude Squad does the same in a terminal with tmux. Anthropic and OpenAI also run parallel sessions in their own apps, each for its own agent only. ### Which alternatives are open source? As of 8 November 2026, by their own repositories: Orca (MIT), Paradev (MIT since 4 November 2026), Emdash (Apache-2.0), Nimbalyst (MIT), Claude Squad (AGPL-3.0), T3 Code (MIT) and Vibe Kanban (Apache-2.0). microManager and Agentastic are free but not open source. ### Which alternative can import my Conductor workspaces? microManager has a one-click, read-only import of Conductor’s repositories, workspaces and chats, and reads Conductor’s settings files and CONDUCTOR\_\* variables. We did not find an import from Conductor described on the other tools’ pages we read on 8 November 2026. ### Can I just use git worktrees without an app? Yes. Run git worktree add for each task and start your agent CLI in each folder. You then handle copying .env files, installing dependencies, picking ports and cleaning up yourself, which is what these apps automate.