FAQ
Day-to-day questions about using Fletch. For questions about isolation guarantees, requirements and beta status, see the main FAQ.
Why macOS only?
Section titled “Why macOS only?”Fletch’s default per-agent isolation uses macOS’s sandbox-exec, which has no equivalent
on other platforms. It requires macOS 13 or newer.
Where are workspaces stored?
Section titled “Where are workspaces stored?”Each agent’s workspace lives under ~/.fletch/workspaces/<id>/, with the repo checked
out in a subdirectory. This directory is separate from Fletch’s app database. (Installs
that predate the current layout may still have old checkouts under ~/.fletch/worktrees/.)
Do I need Docker?
Section titled “Do I need Docker?”No. The default sandbox engine is macOS Seatbelt, which needs no setup. Docker is an optional second engine you can select in Settings › General › Sandbox when you want container-level isolation or Linux builds — see Security & Sandboxing.
Do I need the GitHub CLI?
Section titled “Do I need the GitHub CLI?”No. GitHub features — cloning, creating repos, and pull requests — run on a built-in connection you authorize once in the app (Settings › Account › Connect GitHub). Opening a local folder, spawning agents, committing, and running your project don’t need GitHub at all.
Which agents are supported?
Section titled “Which agents are supported?”Claude Code (the default), Codex, Cursor Agent, and OpenCode are supported; Pi and
Antigravity are experimental. You install and authenticate each CLI yourself (or let
Fletch’s one-click installer do it where available); Fletch detects them on your PATH
and common install locations.
Can I run more than one agent at once?
Section titled “Can I run more than one agent at once?”Yes — that’s the point. Each runs in its own checkout and branch, so any number of agents can work the same repository in parallel without touching each other’s files.
Is the sandbox a full VM?
Section titled “Is the sandbox a full VM?”The default Seatbelt engine is a write-protection sandbox: agents can’t write outside their workspace and their own config directories, but they can still read files and use the network — it’s not air-gapped. The optional Docker engine runs each agent in a Linux container that sees only its mounts. See Security & Sandboxing.
What’s the difference between archive and discard?
Section titled “What’s the difference between archive and discard?”Archive hides a workspace but keeps a restorable snapshot — restore it any time from the History sheet in the titlebar. Discard permanently removes the checkout and deletes its branch.
What’s a fork?
Section titled “What’s a fork?”A new workspace branched from an existing agent, optionally carrying its conversation and even its uncommitted code — see Forking a Workspace.
An agent CLI isn’t detected — what do I do?
Section titled “An agent CLI isn’t detected — what do I do?”Open Settings › Providers, click Re-scan system, and if it’s still not found, set an explicit binary path override for that agent.
Does Fletch upload my code?
Section titled “Does Fletch upload my code?”No. Your code and transcripts stay on your machine — see Privacy.