Archon

Open-source orchestration layer for running multiple AI coding agents in parallel git worktrees.

Free· Free and open source; bring your own model API keysAgentsMulti-model (Claude Code, Codex, others)

Best for

Pick Archon if you want to fan out multiple coding agents across isolated git worktrees and trigger codified workflows from chat or CI.

Skip if

Skip it if you just want a single in-editor AI assistant or aren't ready to self-host and pay for parallel model calls.

Archon is a self-hosted command layer for dispatching AI coding agents like Claude Code and Codex across isolated git worktrees, so several agents can grind on the same repo concurrently without stepping on each other. Workflows are defined once in YAML as directed acyclic graphs with loops and conditions, then triggered from a terminal, web UI, Slack, Telegram, or a GitHub comment.

It is aimed at engineers who have outgrown single-agent IDE plugins and want to fan out tasks (e.g. five bug-fix attempts, a refactor + tests + docs pipeline) across worktrees and then cherry-pick the best result. Pricing is free and open source; you run it on your own machine, bring your own model API keys, and Archon handles the worktree bookkeeping and dispatch surface.

The per-workflow model selection means you can route cheap planning steps to one model and expensive coding steps to another. Caveat: like most worktree-based orchestrators, it inherits the operational tax of managing many concurrent branches and the API spend that comes with running agents in parallel.

Editor's take

Archon is the right shape for teams that have already accepted agents as part of their dev loop and now need to scale them safely. The git-worktree-per-run design is the unglamorous detail that actually makes parallel agents usable. Just be honest about the API bill before you crank up concurrency.

— The AI Tool Bible editorial team