MARTINLOOP

MartinLoop — Give coding agents more work. Watch them less. Ship more.

MartinLoop is the system around the coding agents you already use. It keeps jobs focused, stops waste, checks the work, helps recover when things break, and shows what actually shipped and what it cost. Not another agent. Not another IDE. Built after a $65 autonomous billing incident.

Open source · Apache 2.0 licensed · martin-loop on npm · Keesan12/martin-loop on GitHub

npm install -g martin-loop
martin run "fix flaky CI" --budget 3 --verify "pnpm test"
martin inspect

Give coding agents more work. Watch them less. Ship more.

Agents read repos, edit files, run commands, retry tasks, and produce changes teams may merge. The new question is not whether AI can write code — it is whether the work was measurable, governed, and worth paying for. MartinLoop gives every agent run a budget, verifier, stop rule, and receipt before the change lands.

Not another agent. Not another IDE.

Claude writes. Codex works through the task. Gemini reasons through long context. MartinLoop runs the job around the agents you already use, right in your CLI or IDE — budgets, verification, stop rules, and run records. Use the agent or model you want; the system around the work stays the same.

Agents do the work. MartinLoop governs the work.

  1. Run contract: budget + scope. Define task, dollar budget, timebox, and success criteria before the agent starts.
  2. Policy checks. Validate policies, budgets, allowlists, and risk rules against the run contract.
  3. Agent executes. The coding agent — Claude, Codex, Gemini, or another — plans and acts inside the contract.
  4. Verifier gate. Independent verification checks rules, tests, and invariants before anything is accepted.
  5. Receipt / dossier. Immutable JSONL record of inputs, actions, outputs, and artifacts for every step.
  6. Rollback evidence. Snapshots, diffs, and reversibility data so any run can be undone cleanly.
  7. Stop reason. Every run ends explicitly: success, budget, policy, or error — never silent.
  8. Human review. Final review before release or close, with the full dossier in hand.

FAQ

What is MartinLoop?
The execution control system around AI coding agents. Every run gets a budget, stop rule, verifier, rollback path, and receipt — without changing which agent or model you use.
Is this another coding agent?
No. Claude, Codex, and Gemini are the workers. MartinLoop runs the job around the work they do, right in your CLI or IDE. Nothing to migrate to.
Is MartinLoop open source?
Yes. The core is Apache 2.0 licensed. The hosted dashboard and managed team features are commercial.
Who is it for?
Engineering teams, platform teams, and CTOs running AI coding agents in production who need control, auditability, and a record of what happened.