CLAUDE.md Generator from Repo

Generate a high-quality draft CLAUDE.md for any public GitHub repo in under 30 seconds. Detects stack, common commands, conventions, and structure — no more staring at an empty file when you onboard Claude Code to a new codebase.

What does CLAUDE.md Generator do?

When you start using Claude Code in a new repository, the first thing you should do is write a CLAUDE.md — the project's instruction sheet for the agent. A good CLAUDE.md covers commands, structure, conventions, and "what NOT to do" rules. A bad one (or a missing one) means Claude wastes tokens rediscovering the project on every session.

This Actor:

1. Fetches a public GitHub repo's metadata, README, and manifest files (package.json, pyproject.toml, Cargo.toml, go.mod)
2. Detects the tech stack — languages and frameworks (Next.js, React, Django, FastAPI, etc.)
3. Extracts existing test/build/lint commands from scripts or known tool configs
4. Lists top-level directories so Claude knows where to search
5. Produces a fully-formatted draft CLAUDE.md (saved to the run's key-value store + summary in dataset)
6. Flags if a CLAUDE.md already exists in the repo, so you merge instead of overwrite

The result is a ~70% complete CLAUDE.md that you only need to lightly edit before committing — a 30-minute task compressed to one click.

Why use this Actor?

• Onboard Claude Code to any repo in seconds, not hours
• Standardize team conventions: run this on every project so every CLAUDE.md has the same structure
• Audit your existing repos: see at a glance which projects lack a CLAUDE.md
• Faster contractor / open-source contribution: drop into a new codebase and immediately have the AI agent oriented
• Open-source maintainers: ship a great CLAUDE.md so contributors using Claude Code don't ask repeated "how do I run tests?" questions

How to use CLAUDE.md Generator

1. Enter a public GitHub repo in the form owner/repo (e.g. vercel/next.js)
2. Click Start
3. Wait ~10–30 seconds (depending on repo size)
4. Open the Storage tab → Key-value store → download CLAUDE.md
5. Drop it into the root of your local copy of the repo, edit lightly, and commit

Input

• githubRepo (required) — A public GitHub repository in the form owner/repo. Examples: vercel/next.js, fastapi/fastapi, tokio-rs/tokio.

Output

The Actor saves two outputs per run:

1. Key-value store: CLAUDE.md

The generated draft (Markdown, ~1–3 KB). Sections include:

• Project overview (derived from repo description + README)
• Tech stack (auto-detected)
• Common commands (npm test, cargo build, pytest, etc.)
• Repository structure (top-level dirs)
• Conventions and gotchas (placeholders to fill in)
• What NOT to do (sensible defaults: no npm install without confirming, no commit to main, no claims of "tests pass" without running them)
• Recommended hooks (e.g. verify-before-stop for blocking false-completion regressions)

2. Dataset summary

{
  "repo": "vercel/next.js",
  "stars": 128000,
  "primary_language": "JavaScript",
  "detected_stack": {
    "languages": ["JavaScript/TypeScript"],
    "frameworks": ["Next.js", "React"],
    "testCommands": ["npm test  # or: jest"],
    "buildCommands": ["npm run build  # next build"]
  },
  "existing_claudemd": false,
  "draft_size_bytes": 2148
}

Pricing

Pay-per-event model:

• Per-repo generation: $0.05 per draft (~6 GitHub API calls per run — well under the anonymous rate limit)
• Bulk plan: $5 for 200 drafts (run on every repo in your org in one batch)
• Custom: contact for white-label / on-prem versions

Tips

• The Actor uses anonymous GitHub API (60 req/hour ceiling). For high-throughput org-wide audits, run in small batches or upgrade Apify proxy.
• Generated drafts are a starting point, not a final spec. Review before committing — especially the "conventions and gotchas" section, which is mostly placeholders.
• For repos with an existing CLAUDE.md, the Actor flags it in the output. Merge selectively rather than overwriting.
• For repos missing manifest files, the stack detection falls back to the repo's primary_language field — less precise but still useful.

FAQ

Does this work on private repos?

Not in the public version. The Actor uses anonymous GitHub API. For private-repo support, contact for a custom version that accepts a GitHub token via Apify secrets.

What if the repo uses a stack I haven't listed?

The Actor detects JS/TS, Python, Rust, and Go. If your repo uses another language (PHP, Ruby, Elixir, etc.), the stack section will be blank and you'll need to fill it in. PRs to the source repo welcome.

Is the generated CLAUDE.md good enough to commit as-is?

No — it's a draft. Plan for 5–10 minutes of editing per repo: fill in the conventions, gotchas, and "what NOT to do" sections specific to your codebase. The Actor does the boring 70%; you do the high-value 30%.

How does this compare to writing CLAUDE.md from scratch?

Same final quality if you spend the same time editing. The difference is the floor — without this Actor, most repos end up with no CLAUDE.md at all because the activation energy is too high. With this Actor, every repo gets at least a baseline.

Support

Issues / feature requests: open in the Issues tab on the Apify console.

Built by Ian Mu — github.com/ianymu — author of verify-before-stop, the Stop hook that blocks Claude Code session-end if files changed but no verification log exists.