Docfork MCP Server

@docfork

代码仓库

基本信息

分类

其他

传输方式

stdio

作者

docfork

提交者

Adrian

关于 Docfork MCP Server

Docfork MCP Server - @latest docs of 9000+ code libraries for your AI code editor.

#code #llm

Docfork MCP Server 配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "docfork": {
      "command": "npx",
      "args": [
        "-y",
        "docfork@latest"
      ]
    }
  }
}

Docfork MCP Server 工具

Retrieves up-to-date documentation and code examples for any library. This tool automatically searches for the library by name and fetches its documentation. Usage: 1. Provide the library name (e.g., "react", "express", "mongodb") 2. Specify a topic to focus the documentation on (e.g., "hooks", "routing", "authentication") The tool will: 1. Search for libraries matching the provided name 2. Select the most relevant library based on name similarity, trust score, and documentation coverage 3. Fetch the documentation for the selected library, and the relevant sections by topic Response includes: - Library selection explanation - Comprehensive documentation with code examples - Focused content if a topic was specified

概览

What is Docfork MCP?

Docfork MCP provides the latest documentation for over 9,000 libraries directly inside your AI code editor via a single Model Context Protocol server. It is designed for developers using Cursor, Windsurf, Claude Desktop, VS Code, Zed, Claude Code, BoltAI, or any other MCP-compatible client.

How to use Docfork MCP?

Install Node.js v18 or later, then configure your MCP client. For local use, add a command-based server using npx -y docfork (or bunx -y docfork or deno run --allow-env --allow-net npm:docfork). For remote use, point your client to https://mcp.docfork.com/mcp (SSE for Windsurf/Claude Code). Detailed JSON configuration snippets for every supported editor are provided in the README.

Key features of Docfork MCP

Searches library docs by name and topic with the search tool.
Supports both local (stdio) and remote (SSE) connections.
Works with Cursor, Windsurf, VS Code, Zed, Claude Desktop, Claude Code, BoltAI.
Can be run via Docker for isolated environments.
Accepts a tokens parameter (10,000–50,000) to control response length.

Use cases of Docfork MCP

Get up‑to‑date documentation for a library without leaving your editor.
Focus documentation on a specific topic (e.g., “routing”, “hooks”) to avoid irrelevant content.
Integrate documentation lookups into AI‑assisted code completion and refactoring workflows.
Access docs for thousands of libraries in one unified MCP tool instead of installing separate servers.

常见问题

What are the runtime requirements for Docfork MCP?

Node.js v18 or later is required. For clients that support it, you may also use Bun or Deno as alternative runtimes.

Can I use Docfork MCP remotely without installing anything locally?

Yes. You can configure your MCP client (e.g., Cursor, Windsurf, VS Code, Claude Code) to connect to the remote endpoint `https://mcp.docfork.com/mcp` (or `https://mcp.docfork.com/sse` for SSE‑based clients).

What tool does Docfork MCP expose?

It exposes a single tool called `search`. It requires `libraryName` (the library to look up) and `topic` (to focus the docs). Optionally you can set `tokens` (default 10,000, max 50,000) to limit the response size.

How can I troubleshoot common installation errors?

If you see `ERR_MODULE_NOT_FOUND`, try using `bunx -y docfork` instead of `npx`. For `ESM` resolution errors, add the `--experimental-vm-modules` flag. You can also append `@latest` to the package name to force the newest release.

Is the documentation from Docfork MCP guaranteed to be accurate?

No. Docfork is a community‑driven catalogue. While submissions are reviewed, the project makes no warranties—express or implied—about the accuracy, completeness, or security of any linked documentation or code.

其他分类下的更多 MCP 服务器

ICSS

chokcoco

不止于 CSS

20.7K

MCP Toolbox for Databases

googleapis

MCP Toolbox for Databases is an open source MCP server for databases.

15.9K

Glasswarp

See and control a real Windows PC you own — from any MCP client, locally or remotely. Observe (UIA + screenshots), click/type/drag/scroll, launch apps, owner Live View. BYOH: your machine, your key.