Worktrunk

Worktrunk is a Git worktree command-line tool aimed at running several AI coding agents in parallel. The official site is https://worktrunk.dev and the source is https://github.com/max-sixty/worktrunk . GitHub describes it as a CLI for Git worktree management, designed for parallel AI agent workflows. At the time of this scan the repository had about 12,295 stars; it was created on 17 October 2025. It was still on GitHub's weekly trending list in this window.

Git worktrees let you check out more than one working directory from the same repository. Worktrunk wraps that into a CLI for agents: each task or agent can get its own worktree, so they are less likely to overwrite one shared dirty working tree. Exact subcommands, install steps, and how it sits next to tools such as Claude Code or Codex should be taken from the site and README as they stand when you use it.

Highlights:

• Site: https://worktrunk.dev ;
• Repository: https://github.com/max-sixty/worktrunk ;
• It is for parallel agents or feature branches, not a replacement for Git itself.

Why it matters

When several coding agents edit one repo at once, the usual failure is a shared working tree: uncommitted work gets clobbered and branches interrupt each other. Worktrees are already in Git, but the default commands are awkward if you just want to hand an agent a clean tree and delete it afterwards. Worktrunk packages that workflow as a dedicated CLI and says so in its description, which is why the name is worth recording as a product, not only as a Git tip.

It has also stayed on GitHub trending, which is a sign that people are actually running multi-agent work against one repository. The previous radar slot verified it and held it for capacity; this slot adds the page.

Limits

Worktrunk isolates worktrees and branches. It does not, by itself, make agent-written code correct, and it does not replace tests, review, or permissions. Star counts change; use the GitHub page at the time you look. Install steps, default branch rules, and integrations with a given agent product must come from the site or README — do not assume undocumented automation. Opening many worktrees on a shared machine also has disk and hook-script cost.

1. Open https://worktrunk.dev or https://github.com/max-sixty/worktrunk and install the CLI from the current README;
2. In a real repository, give each parallel task its own worktree so agents or your own sessions work in separate directories;
3. Remove worktrees when the task is done, following the docs, so leftover branches and disk use do not pile up;
4. Use it together with review, tests, and agent permissions — worktree isolation is not a security boundary on its own.