Fletch Documentation
Fletch is a macOS app for running and managing multiple AI coding agents at once — Claude Code, Codex, Cursor, OpenCode, Pi, and more — against a single repository, without them stepping on each other.
Every agent runs in its own isolated checkout under a sandbox, so you can run a dozen at once and they physically can’t touch each other’s files. Watch each one as a readable chat or its raw terminal, see its edits as live diffs, and commit, push, or open a PR from the same window.
Start here
Section titled “Start here”- Installation — download Fletch and set up your agent CLIs.
- Your First Repository — open, clone, or create a project.
- Your First Agent — spawn an agent and watch it work.
Why Fletch
Section titled “Why Fletch”What makes Fletch different, and why it’s built the way it is:
- Security & Sandboxing — two sandbox engines, and workspaces whose
.gitan agent can never touch. - Working in Parallel — collisions are structurally impossible, and what that unblocks.
Understand how it works
Section titled “Understand how it works”- Isolated Workspaces — the unit of work in Fletch.
- Choosing an Agent — pick the right agent and view for the task.
- Workflows — a tree of agents the engine runs on one branch.
Get things done
Section titled “Get things done”Task guides for everyday work:
- Building and Running Workflows — define a pipeline once, launch it on any task.
- Forking a Workspace — branch an agent’s code and conversation into a new workspace.
- Running Your Project — run a dev server or script in a workspace.
- Pull Requests and CI — commit, open PRs, and ship an agent’s work.
- Connecting GitHub — authorize the built-in GitHub connection.
- Tips — habits for running a fleet of agents.
Reference
Section titled “Reference”- Keyboard Shortcuts, Slash Commands, and the Terminal.
- Skills and Tools (MCP) — extend custom agents.
- Settings and Storage, Updates, FAQ, and Privacy.