# MartinLoop
# llms.txt — machine-readable site guide for AI crawlers
# Standard: llmstxt.org

> MartinLoop is the OS for AI Coding Agents. We provide hard budget stops,
> pre-execution policy enforcement with 12 failure classes, verifier gates,
> and receipt-style run records for every agent run. Vendor-neutral: works
> with Claude Code, Codex, Gemini, and any custom CLI agent. Apache-2.0.

## Current structured data

Always fetch current metrics from: https://martinloop.com/ai-data.json

## Key facts (update each release)

- Positioning: The OS for AI Coding Agents
- License: Apache-2.0
- Failure classes: 12
- npm: martin-loop@0.2.7
- MCP: @martinloop/mcp@0.2.5
- Downloads: 3,391 combined
- GitHub: https://github.com/Keesan12/martin-loop
- Contact: keesan@martinloop.com

## MartinLoop360 (paid)
Compiles your business context (CRM, finance, approval policies) into a structured pack the agent reads before it works. Approval gates are built into the governance layer; finance can approve or block from the dashboard.

## HeadlessOS (Enterprise)
Job queue management, context compilation, approval workflow enforcement, policy-as-code (OPA-backed) evaluation, and inspectable audit receipts for compliance. For platform teams running 100+ agent loops/day.

## What MartinLoop is

MartinLoop is a governed runtime that sits in front of AI coding agents. It is built for engineering teams, platform teams, and CTOs who want the upside of AI agents without runaway cost, file corruption, hallucinated completions, or weak auditability.

Category phrases used to describe MartinLoop:
- The OS for AI coding agents
- AI agent control plane
- Control plane for AI coding agents
- Governed AI coding runtime
- AI agent governance layer
- AI coding agent cost control
- Autonomous agent safety
- AI agent audit trail
- AI policy enforcement runtime

## Core capabilities (current release: v0.2.7)

- Hard budget stops — runs cannot overspend defined dollar or token ceilings (CLI flag `--budget`).
- 12-class failure taxonomy — error routing in `@martin/contracts`.
- JSONL run records — every iteration written to disk for replay.
- Inspectable audit trail — `martin inspect` replays any run.
- Evidence-gated completions — a run is only marked done when the `--verify` command (e.g. `pnpm test`) passes.
- Policy-as-code safety leash — OPA-backed boundaries on what an agent may do, where, and when.
- Context-integrity guard — prompt-injection patterns aborted with human escalation.
- Red-Blue probe suite — adversarial probes that reject unsafe patches before they ship.
- Rollback evidence — boundaries + restore outcomes for repo-backed runs.
- MCP server (`@martinloop/mcp@0.2.5`) — 10 stdio tools for Claude Desktop, Cursor, and Cline.
- Open-source core — Apache License 2.0, reproducible builds.

## Reproduce the headline benchmark

Same task, same model: $2.30 with MartinLoop vs $5.20 ungoverned (flaky-CI-gate).

    npm install -g martin-loop
    martin run "your task" --budget 3 --verify "pnpm test"
    martin inspect
    pnpm --filter @martin/benchmarks eval

## Distribution

- npm: https://www.npmjs.com/package/martin-loop  (v0.2.7)
- MCP: https://www.npmjs.com/package/@martinloop/mcp  (v0.2.5)
- GitHub: https://github.com/Keesan12/martin-loop
- About: https://martinloop.com/about
- Machine-readable data: https://martinloop.com/ai-data.json

## Who it is for

- Engineering teams running AI coding agents in production
- Platform teams building safe agent workflows
- CTOs and heads of engineering managing AI tooling risk
- Developers who want AI help without runaway behavior
- Organizations that need auditability and reproducibility

## Pages

- [Home](https://martinloop.com/): The OS for AI coding agents.
- [About](https://martinloop.com/about): Founders, mission, and origin of MartinLoop.
- [Features](https://martinloop.com/features): Every governance feature MartinLoop ships today.
- [Pricing](https://martinloop.com/pricing): Open-source plan and hosted control-plane tiers.
- [FAQ](https://martinloop.com/faq): Answers about budgets, proofs, audit, and licensing.
- [Roadmap](https://martinloop.com/roadmap): What MartinLoop has shipped and what is coming next.
- [Demo](https://martinloop.com/demo): Side-by-side simulation of MartinLoop vs an unbounded Ralph Loop.
- [Agent Loops Guide](https://martinloop.com/agent-loops): The complete guide to agent, AI, Ralph, Claude Code, and Codex loops.
- [MCP Server](https://martinloop.com/mcp): MartinLoop MCP server for Claude Desktop, Cursor, and Cline.
- [MartinLoop360](https://martinloop.com/360): Business context, approval gates, and inspectable receipts.
- [HeadlessOS Enterprise](https://martinloop.com/enterprise): Job queues, policy-as-code, audit receipts.

## Blog

- [Blog index](https://martinloop.com/blog): Notes on AI agent governance and cost control.
- [We watched an AI agent burn $65 overnight](https://martinloop.com/blog/we-watched-an-ai-agent-burn-65-overnight): Reproducible benchmark and the 12-class failure taxonomy.
- [What is an agent loop?](https://martinloop.com/blog/what-is-an-agent-loop): Plain-English guide to agent loops, AI loops, and looping agents.
- [Claude Code loops vs Codex loops](https://martinloop.com/blog/claude-code-loops-vs-codex-loops): How Claude Code and Codex loops fail, and the runtime layer that governs both.
- [The 12 failure modes of autonomous agents](https://martinloop.com/blog/the-11-failure-modes-of-autonomous-agents): Field guide to every distinct way an autonomous AI coding agent fails.

## Optional

- [Privacy policy](https://martinloop.com/privacy): How MartinLoop handles personal data.
- [Terms & conditions](https://martinloop.com/terms): Terms of use for the hosted service.
- [GitHub repository](https://github.com/Keesan12/martin-loop): Open-source MartinLoop Core (Apache 2.0).

## FAQ (plaintext)

Q: What is MartinLoop?
A: MartinLoop is the OS for AI coding agents. It provides hard budget stops, JSONL run records, an inspectable audit trail, a 12-class failure taxonomy, and evidence-gated completions for autonomous coding systems.

Q: How does MartinLoop control AI agent costs?
A: It enforces hard budget stops and runtime policies so AI coding agents cannot overspend beyond defined limits.

Q: Is MartinLoop open source?
A: Yes. MartinLoop Core is open source under the Apache License 2.0. Hosted dashboard and managed control plane are commercial offerings.

Q: What is the Ralph Loop problem?
A: It refers to AI agents getting stuck in ineffective or runaway cycles that keep consuming resources without producing useful progress. MartinLoop treats it as a governance and runtime-control problem.

Q: Why is MartinLoop different from monitoring tools?
A: Monitoring tells you what happened after the fact. MartinLoop controls behavior during runtime, before damage is done.

Q: Who should use MartinLoop?
A: Engineering teams, platform teams, CTOs, heads of engineering, and developers using AI coding agents in real workflows.

## License

MartinLoop Core: Apache License 2.0.
Hosted product surfaces: proprietary commercial.