GitHub radar
Foreman is an independent supervisor for Codex programming agents
Foreman is a Python runtime that works in tandem with Codex coding agents and can determine if code is finished, if the agent is stuck, or if a human should intervene.
Foreman is a Python runtime that can work with Codex coding agents to tell when code is complete, when an agent is stuck, or when human intervention is needed. Foreman tracks the progress of a software development task while a coding agent is working. Every 30 seconds it looks at the git diff, agent logs, and step history to ask a fast classifier 9 yes/no questions about whether the code is implemented, there are enough tests, the agent is stuck, the code is veering off track, or human intervention is needed. A deterministic Python policy is then used to decide whether to proceed, steer, stop, retry, check, or escalate. Foreman is an architectural experiment — not for production. Calibrate accuracy. Requires Codex CLI & TypeSafe API Key or run foreman demo --repo . locally.
Why a vibe-coder should care
Coding agents can get stuck or produce poor results on long running tasks. Use Foreman as a second pair of eyes to monitor the progress of your project and avoid your agent from getting stuck (and wasting tokens) or producing errors.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Deploy Foreman from https://github.com/thruwire/foreman — follow the README, start with the offline demo `foreman demo --repo .` (no keys needed); ask me for TypeSafe API and Codex credentials if you want to run a real job
Runs on any laptop — needs Python, Codex CLI, and a TypeSafe API key
Open on GitHub▌ More finds