Claude Code Docs workflow
An archive of Anthropic documentation in Markdown: Claude Code, API, MCP, products, support, 4100 files, updated four times daily
- 172 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The agent reads local archive files and there are no network actions when reading
- For a self-run refresh the fetcher visits Anthropic sites and GitHub and reads your token from the environment
Install
- Works in
Manual install
Run `git clone https://github.com/thevibeworks/claude-code-docs` and enter the folder. Open Claude Code in that folder and answer questions from the files in `content/`, naming the file path. To update use `uv run scripts/fetcher.py --incremental`.
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The agent reads local archive files and there are no network actions when reading
- For a self-run refresh the fetcher visits Anthropic sites and GitHub and reads your token from the environment
This is third-party code. Review the repository files before installing.
What it does
The repository keeps a copy of Anthropic's developer materials that CI refreshes four times a day. You clone it and open it in Claude Code: the agent answers questions from local files, search works offline and version history lives in Git. The archive has six live sources: Claude Code and Agent SDK docs (198 files), the API reference (2228), products (226), the MCP specification (347), Anthropic repositories on GitHub with cookbooks, skills and plugins (764) and support (372). The anthropic.com blog has been frozen since July 8, 2026: engineering and product posts are kept in full, research at 46 percent and news at 22. The `scripts/fetcher.py` script on uv downloads sections again (`--section`, `--incremental`, `--tree`, `--discover`), and every source serves a Markdown version, so nothing is converted from HTML.
Who it is for. Developers building on Claude Code, the API and MCP who want the agent to answer from current documentation in local files, without relying on model memory.
Good fit when
- The agent is confused about hooks or subagent settings and an answer is needed from fresh documentation
- You want to find the description of tool annotations in the MCP specification without going online
- You want to reread the engineering article "Building Effective Agents" from a local copy
- You need to trace what changed in the documentation between dates using Git history
Not a fit when
- The very latest version of a page is always needed: the archive can lag between refreshes
- The full anthropic.com research and news posts are needed: the archive covers them partly and is frozen
- Disk space is limited: the archive has about four thousand files
Example request
How do I set up hooks in the Agent SDK according to this documentation? Show the file the answer comes from.Limitations
The `content/blog/` section has been static since July 8, 2026, because anthropic.com serves only HTML and the jina.ai proxy the fetcher used was removed in July 2026: research is covered at 46 percent and news at 22. File counts are given as of September 9, 2026. Updating by yourself needs uv, and the `github` section needs a `GITHUB_TOKEN` or `GH_TOKEN` token in the environment. The content belongs to the authors of the sources and the repository keeps copies. The latest commit in the repository is from November 2026.
How to disable. Delete the repository clone: it changes no Claude Code settings. If you opened it as a project folder, close the session and remove the folder from disk.
FAQ
How do I use the archive?
Clone the repository, enter the folder and run `claude` with a question, for example `claude "how does the agent loop work?"`. The agent looks for the answer in local files.
What does the archive contain?
Claude Code and Agent SDK docs, the API reference, products, the MCP specification, Anthropic repositories on GitHub, support and a frozen blog with engineering articles.
How often is it updated?
Four times a day, by the `fetch-claude-docs.yml` workflow. For a manual refresh there is `uv run scripts/fetcher.py` with the `--section` and `--incremental` flags.
Why is the blog frozen?
anthropic.com serves only HTML and the jina.ai proxy through which the fetcher got Markdown was removed in July 2026. The blog archive stopped on July 8, 2026.
Is a GitHub token needed?
Only for refreshing the `github` section: the fetcher reads `GITHUB_TOKEN` or `GH_TOKEN` from the environment. Reading the ready archive needs no token.
Discussion
No questions or reviews yet.
Related
WorkflowsA Claude Code plugin that generates and maintains CLAUDE.md: a 0 to 100 quality score, a 150-line cap and stale reference checks
Anthropic DevRel playground: three Claude Code mods, hooks, CLAUDE.md examples and skills, each with notes on how it was built.
An MCP server that gives an agent search over a large Claude Code guide, official docs, releases and ready-made templates
Claude Code plugins: plan, work and review commands, subagents, hooks and writing skills in a practical minimalism spirit.
Claude Code Blueprint: 12 agents, 19 skills, 16 hooks and 6 CLAUDE.md rules, copy what you need or install a preset
A Claude Code working configuration: a planner with per-milestone review, sub-agents and skills for analysis, refactoring and docs sync
README in short
The README explains how to use the archive with Claude Code and gives a table of sources with file counts and fetcher sections. The frozen blog, the `content/` folder structure and update commands are described separately.