Skip to main content
Glama

🩺 repo-ready

The doctor your repo needs before AI agents move in.

One command generates a truthful AGENTS.md + CLAUDE.md from what's actually in your repository β€” commands proven by CI, real directory map, env vars, guardrails, definition-of-done. A second command audits them for drift, so your agent docs never rot again.

License: MIT Node Zero dependencies Works with

npx repo-ready init Β· npx repo-ready check Β· ships as CLI, Agent Skill and MCP server

Why Β· Quick start Β· Check mode Β· MCP & Skill


Why

Every AI coding agent asks the same first question: "what is this project and how do I run it?" Today you answer it by hand-writing AGENTS.md β€” and then it silently rots: scripts get renamed, CI changes, folders appear.

repo-ready fixes both ends:

  1. init β€” generates agent docs from evidence: manifests, lockfiles, CI workflows, Makefiles β€” with a confidence column. "verified (manifest + runs in CI)" beats a hallucinated guess every day.

  2. check β€” audits existing docs against the live repo: dead commands, undocumented scripts, new directories, CI drift. Exit code 1 = unhealthy β†’ drop it into CI.

  βœ… Repo health: HEALTHY (100/100)

   βœ… repo-ready managed (v0.1.0)
   βœ… command OK: `npm run test`
   ⚠️  package.json has scripts not documented in AGENTS.md: `db:migrate`
   ❌  documented command `npm run ghost` no longer exists in package.json scripts

Related MCP server: MDPilot MCP

Quick start

Zero dependencies, zero config:

cd your-repo
npx repo-ready scan      # see what it detected β€” no files written
npx repo-ready init      # generate AGENTS.md + CLAUDE.md
npx repo-ready check     # audit any time (CI-friendly exit codes)
npx repo-ready update    # regenerate after the project evolved

What init writes (example from a real repo):

## Commands

| Purpose | Command | Confidence |
|---|---|---|
| install  | `npm install`    | verified (package.json + lockfile) |
| dev      | `npm run dev`    | verified (package.json scripts)    |
| test     | `npm run test`   | verified (manifest + runs in CI)   |

## Repository layout

src/ β€” application source code
public/ β€” static assets served as-is
scripts/ β€” utility / maintenance scripts

## Environment variables
- `DATABASE_URL` β€” no default, must be set

## Do not touch
- `dist/` is generated β€” never edit or commit it
- `.env` holds real secrets β€” never read it, commit it, or echo it

## Definition of done
- Run `npm run test` and make it pass before claiming a task is done
- Do not commit unless the user asked for a commit

Hand-written notes survive. Everything repo-ready owns sits between <!-- repo-ready:auto-start --> markers β€” your own paragraphs above and below them are kept on every update. A foreign (hand-written) AGENTS.md is never overwritten without --force.

What it detects

Languages

JS/TS, Python, Go, Rust, Java, Ruby, PHP, C/C++, Vue, Svelte, … (extension census)

Package managers

npm / pnpm / yarn / bun, poetry / uv / pipenv / pip, cargo, go modules

Frameworks

Next.js, React, Vue, Express, Fastify, NestJS, FastAPI, Django, Flask, Gin, axum, …

Commands

from package.json scripts, pyproject.toml, go.mod, Cargo.toml, Makefile

CI evidence

parses .github/workflows and upgrades commands to β€œruns in CI”

Guardrails

generated dirs, lockfiles, .env, published-package versioning

The check mode 🩺

npx repo-ready check && echo "docs are honest" || echo "docs rotted"

Verifies that every documented command still exists, flags scripts/dirs missing from the docs, compares against what CI actually runs, and scores the result 0–100 (healthy β‰₯ 90, stale, critical). Non-zero exit on anything but healthy β€” wire it into CI and your agent docs can never silently rot again.

Plug into your agent

Add this repo as a plugin marketplace and install β€” you get the skill, a /repo-ready command and the MCP server in one shot:

  • ZCode: Plugin Marketplace β†’ Add β†’ paste temmeik/repo-ready β†’ install Repo Ready

  • Claude Code: claude plugin marketplace add temmeik/repo-ready β†’ /plugin install repo-ready@repo-ready

As an Agent Skill (manual)

Copy skills/repo-ready/SKILL.md into your skills directory: ~/.claude/skills/repo-ready/SKILL.md β€” then just ask the agent to "make this repo agent-ready".

As an MCP server

# Claude Code
claude mcp add repo-ready -- npx github:temmeik/repo-ready mcp

# Codex (~/.codex/config.toml)
[mcp_servers.repo-ready]
command = "npx"
args = ["github:temmeik/repo-ready", "mcp"]

Tools exposed: repo_scan Β· repo_generate Β· repo_check

Monorepo? Rust? Python?

Yes. Workspaces, Turborepo and pnpm-workspace are detected; Go/Rust/Python get native command sets (go test ./..., cargo test, poetry run pytest) β€” all covered by 41 e2e assertions in test/e2e.js, including a run against this repo's own sister project.

License

MIT β€” do whatever you want, a star is appreciated ⭐

Related MCP Connectors

Related MCP Servers

  • A
    license
    Not graded
    quality
    D
    maintenance
    Provides documentation generation and analysis tools for AI agents, including JSDoc generation, README analysis, changelog generation, API documentation, and code comment analysis.
    19 npm
    MIT
  • F
    license
    Not graded
    quality
    D
    maintenance
    CP server that generates AGENTS.md and CLAUDE.md from your real repo, verifies every command actually exists, detects doc drift, and saves session context β€” all locally, no data leaves your machine. Free-tier friendly via Groq/NVIDIA.
    -
  • A
    license
    A
    quality
    A
    maintenance
    Checks the context files a coding agent reads (CLAUDE.md, AGENTS.md, SKILL.md, .cursor/rules) against the git index of a repository, and reports wrong letter case, broken links, missing paths, commands nothing defines and agent configuration that points at nothing. Two more tools report which sections drifted from the code and what each file costs; zero dependencies, stdio, Node 18 or newer.
    5
    80 npm
    5
    MIT