What is an agent-ready codebase?

A repo AI coding agents can work in safely: an AGENTS.md rules file, a one-command verify script, CI, tests, and security guardrails like .env.example and RLS-style conventions. The Agent-Ready Score grades all of it out of 100.

Is the Agent-Ready Score free?

Yes — completely free, no sign-up. It reads public GitHub repos via the public API and grades them instantly.

What does the score check?

Five categories: agent rules (AGENTS.md, CLAUDE.md, .cursorrules), verification (CI, tests, verify script), guardrails (.env.example, lockfile, license), docs, and typed structure.

Free tool · no sign-up

Is your repo ready for AI agents?

Paste any public GitHub repo. We run a pre-flight inspection — agent rules, verification, guardrails, docs — and hand you a graded scorecard out of 100.

Try it on vercel/next.js or your own repo.

Agent rules · 30 pts

AGENTS.md at the root (the Linux Foundation-stewarded standard read by 28+ tools), with real substance and cross-tool coverage for Claude Code, Cursor, and Copilot.

Verification · 25 pts

CI pipelines, tests, and a one-command verify script — the difference between an agent that checks its work and one that just claims success.

Guardrails · 20 pts

.env.example, .gitignore, a committed lockfile, and a license — the hygiene that stops an agent leaking secrets or breaking installs.

Docs · 15 pts

A real README, a contributing guide, and a docs folder — what agents read before they start inventing answers.

Typed structure · 10 pts

Typed-language foundations so the compiler catches what code review misses.

Want an A without the homework?

AgentShip ships scored A+ out of the box — or start free with AgentShip Lite.

116 repos scored

Browse popular repos

Open any repo below to see its live Agent-Ready scorecard — how well it's set up for Claude Code, Cursor, and Copilot.