Claude Code Docs plugin
Claude Code Documentation Tool
A Claude Code plugin that keeps the official Claude docs at hand, reading each page live from Anthropic on demand, with no mirror and no MCP server
- 58 ★ on the repository
- Last commit
Extended access
Why this level
- A session-start hook syncs metadata via git and fetches pages from the network
- Runs a bash client that downloads files into a local cache
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/claude-code-documentation-toolDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- A session-start hook syncs metadata via git and fetches pages from the network
- Runs a bash client that downloads files into a local cache
This is third-party code. Review the repository files before installing.
What it does
The plugin gives Claude Code a /docs command and an auto-discovery skill that reads the relevant official Claude doc pages before answering. The repository holds only metadata, about 3 MB: a page manifest and a prose-free search index, while the pages themselves are fetched directly from Anthropic's domains into a local cache. It searches the index first and reads only the pages it needs, so little extra lands in context. It also generates interactive HTML courses on any topic and changelog reports for a chosen period. On session start the metadata updates and changed pages re-fetch in the background.
Who it is for. For people working in Claude Code who want accurate answers about Claude, the API, and the SDK with links to current doc pages.
Good fit when
- You want an answer about Claude Code, the API, or the SDK with a link to the current page
- You want the agent to read the docs itself before answering
- You need a course or a changelog report on the documentation
Not a fit when
- You are not working in Claude Code or a compatible agent
- You need documentation for products other than Anthropic's
Example request
How do I set up MCP servers in Claude Code? Answer from the current docs with linksLimitations
The client runs on macOS and Linux and needs jq, curl, and bash. Pages are read directly from Anthropic's domains platform.claude.com and code.claude.com, whose access from Russia without a VPN is not guaranteed, so doc fetching may fail. On session start a hook hard-syncs the metadata via git reset and fetches pages in the background.
How to disable. Disable the plugin with /plugin uninstall claude-docs@claude-code-docs. The local cache lives in ~/.claude-code-docs and can be removed there.
FAQ
Does the plugin copy Anthropic's docs?
No. The repository holds only metadata, a manifest and a prose-free search index. The pages themselves are read from Anthropic's servers on demand.
What dependencies are needed?
The client is bash, curl, and jq and runs on macOS and Linux. No separate server, Python, or API keys are required.
Discussion
No questions or reviews yet.
Related
PluginsPlugin for Claude Code and Codex: the agent answers with an interactive page of charts and calculators that you can edit and keep.
Atlassian Confluence CLI with a Claude Code plugin: read, search and edit pages from the terminal, with a read-only agent mode.
Multi-agent documentation generator: analyzes a repository and builds a README and AI assistant config files
A plugin for studying research papers: a PDF or arXiv link becomes a summary, study materials, code demos and a web viewer
A Claude Code plugin that writes papers, grants and posters from real literature: source search first, then cited text.
explainmyrepo turns a GitHub repository into an explainer page and a knowledge pack an agent answers from.
README in short
The README introduces a plugin that keeps the official Claude docs current inside Claude Code. It installs in two /plugin commands, provides a /docs command and an auto-discovery skill, plus an interactive course generator and changelog reports. The architecture stores only metadata, about 3 MB, and reads pages directly from Anthropic's domains without copying anything into the repo. The client is bash, curl, and jq, and updates run through a session-start hook and GitHub Actions every three hours. It has a fetch safety layer with a domain allowlist and checks, an MIT license, and an automated test suite.