herdr-rig
One manager agent, many coding agents, one terminal workspace.One manager agent, many coding agents, one terminal workspace. You talk to one session; it does the juggling.
What it does for you.
You talk to one session.
The manager turns a request into tickets and hands them out.The manager turns a request into tickets with a goal, checks and verify commands, then hands them out.
Any harness, any role.
Claude Code, Codex, opencode, pi, omp or Command Code, each with its own model.Claude Code, Codex, opencode, pi, omp or Command Code, each with its own model. A cheap worker, a reviewer from another family.
Nothing to babysit.
Own worktree per ticket. A watcher speaks up when an agent goes quiet.Each ticket gets its own worktree. Workers report back on their own, and a watcher speaks up when one goes quiet.
Why a manager?
Running several agents at once is easy. Keeping track of them is not, so I gave the job to one agent.Running several agents at once is easy. Keeping track of them is not: which one is stuck on a question, did the review pass, whose branch is safe to merge. I kept losing that thread, so I gave the job to one agent.
What’s tested.
TestedWorkers, all six harnesses
TestedWatcher, board and cleanup
Not yetmerge pr and auto-merge
WelcomeReports from macOS and Linux
- Workers, all six harnessesTested
- Manager in the herdr desktop appTested
- Watcher, board and cleanupTested
- merge pr and auto-merge, end to endNot yet
- Driving panes by hand on macOS and LinuxReports welcome