DevSwarm is a desktop app for running multiple AI coding assistants in parallel, each isolated on its own Git branch. This guide covers installation and your first workspace.
- macOS or Windows (Windows includes WSL repository support)
- Git available in your
PATH - At least one CLI-based AI assistant installed — DevSwarm detects them automatically. See Supported AI Assistants.
- Download the installer from devswarm.ai/download.
- Run the installer for your operating system.
- Launch DevSwarm and sign in with Google or GitHub.
After you sign in, DevSwarm opens the home page. If you have no repositories yet, a welcome screen walks you through adding your first one.
Add a repository by importing one from your local filesystem, cloning from a remote URL, cloning from GitHub via OAuth, or creating a new one. Set the Default Branch — DevSwarm uses it as the source branch when it creates workspaces.
Tip: Start from a clean state (everything committed) on the branch you set as default.
A workspace is an isolated Git worktree paired with an AI assistant terminal and a code editor. When you create one, DevSwarm:
- Creates a new branch and worktree from the source branch.
- Starts your chosen AI assistant's CLI in that worktree.
- Opens Build Mode with an active AI terminal session.
You can also create workspaces directly from a GitHub pull request or Jira issue — DevSwarm hands the PR or issue context to the assistant automatically.
- Build Mode (
Cmd+1/Ctrl+1) — a code editor alongside tabbed AI and shell terminals, with file-explorer and Git panels. - Review Mode (
Cmd+2/Ctrl+2) — a diff viewer showing your workspace branch against its source branch, with syntax highlighting. - Open the workspace in an external editor (Cursor, JetBrains, and others) with
Cmd+5/Ctrl+5.
Create more workspaces to run assistants in parallel — a refactor in one, tests in another, docs in a third. Each has its own branch, filesystem, and port assignments, so nothing collides.
Questions or bugs? Open an issue or email support@devswarm.ai.