Claudio

Claudio runs on your machine and coordinates multiple Claude Code sessions working on the same repository. It's for developers who want to split coding work across parallel agents while keeping each agent's edits on a separate branch. It requires an authenticated Claude Code CLI, Git and tmux.

Each session gets its own git worktree, so agents can change files independently. Shared context tells them what other sessions are doing, which files they're editing and how their tasks relate. Task dependencies let work wait for prerequisites rather than starting everything at once.

The terminal dashboard shows session status and streams agent output. You can inspect diffs, search output and pause or stop individual sessions. Claudio also recovers sessions after a disconnection and tracks token usage and API costs with configurable limits.

Planning goes beyond dividing up a task manually. Claudio can analyze a codebase and produce a structured plan, then coordinate parallel execution through UltraPlan. Multi-pass planning compares competing strategies. TripleShot runs several attempts at the same task and uses a judge to choose one, while adversarial review pairs implementation with repeated critical feedback.

For work headed to GitHub, Claudio can generate pull requests and assign reviewers through the GitHub CLI. Session state, worktrees and logs live in the local project directory; coding requests run through the authenticated Claude Code backend.