Report Abuse

Basic Information

This repository provides an open source MCP (Model Context Protocol) server that exposes llms.txt documentation sources to MCP-capable host applications and IDEs. It is designed to let developers supply a user-defined list of llms.txt files and to offer simple tools such as list_doc_sources and fetch_docs so host apps can request documentation URLs and read their contents. The server is intended to increase transparency and auditability of tool calls and the context returned to an LLM-driven app. The project includes a command-line interface, sample YAML/JSON configs, a programmatic create_server API, and integration examples for Cursor, Windsurf, Claude Desktop, and Claude Code. It runs via the uv/uvx runtime and supports transports like stdio and SSE. Security controls require explicit allowed domains for remote and local llms.txt usage to prevent unauthorized network access.

Links

Categorization

App Details

Features
Provides two primary tools for MCP hosts: list_doc_sources to enumerate configured llms.txt files and fetch_docs to read URLs referenced by those files. Strict domain access controls automatically add domains for remote llms.txt but require explicit --allowed-domains for local files; --allowed-domains "*" is supported. CLI supports --yaml, --json and repeated --urls name:url entries, plus options like --follow-redirects and --timeout. Includes sample_config.yaml and sample_config.json, programmatic create_server function for embedding, and examples of how to register the MCP server with Cursor, Windsurf, Claude Desktop and Claude Code. Supports transports like stdio and SSE and can be run locally for testing with tools such as the MCP inspector. Uses uv/uvx for runtime invocation.
Use Cases
MCPdoc gives developers control and visibility over documentation tool calls that IDEs and apps make when they consume llms.txt files. By centralizing llms.txt sources behind an MCP server, teams can audit each fetch_docs invocation, inspect returned context during debugging, and limit network access via domain whitelists. The repo supplies CLI and programmatic options so it can run as a local service during development or be embedded in workflows. Integration guides and sample configs make it straightforward to register the server with popular host apps, enabling developers to require explicit approval for tool calls and to combine multiple documentation sources for richer, auditable context.

Please fill the required fields*