cowork
Creator and maintainer
Open-source Python CLI that orchestrates multiple AI coding agents into role-based teams, with phase gates, per-role evals and cost reporting.
- python
- ai-agents
- llms
- observability
What it is
cowork is an open-source Python CLI I built and maintain for orchestrating multiple AI coding agents — Claude Code, Codex and opencode among them — into role-based teams (scout, planner, builder, reviewer) working a task through phase gates, rather than one agent freelancing the whole thing.
What I built
- A role-based orchestration model with phase gates between scout, plan, build and review, so a team of agents cannot skip straight to "build" on an unscoped task.
- Per-role evaluations, so agent output is graded against the role it was playing, not judged by a single blanket pass/fail.
- Cost reporting per session and per role, on the same logic behind the evaluation-infrastructure work I do professionally: an AI feature is accountable to its cost, not just its output.
Why it mattered
It is the clearest evidence I have of building my own tools rather than only consuming other people's, and of shipping outside the TypeScript/Node stack that dominates my day job — cowork is a Python project, end to end.