Skip to content
MCP Server Directory logo
developer-tools

Server Llm Txt

Archived: the mcp-get community server registry is no longer maintained.

Reviewed overview

MCP 服务器概览

What is the project about?

Server Llm Txt is an MCP (Model Context Protocol) server that extracts and serves context from llm.txt files. It exposes the LLM.txt Directory, allowing clients (e.g., Claude Desktop or other MCP-aware tools) and AI models to list available llm.txt files, fetch file contents, and run multi-query searches with contextual snippets. Note: the project is archived and the mcp-get community server registry is no longer actively maintained.

How to use the project?

  • Recommended (one-liner): install via MCP Get so it is automatically configured in Claude Desktop: npx @michaellatman/mcp-get@latest install @mcp-get-community/server-llm-txt
  • Manual configuration: add an MCP server entry to your claude_desktop_config.json that runs the server package via npx (example provided in repo).
  • Development: npm install && npm run watch to run with automatic recompilation.
  • Runtime behavior: the server provides RPC-style tools (list_llm_txt, get_llm_txt, search_llm_txt) for clients to call; results are cached locally for 24 hours in OS-specific cache locations.

Key features

  • Directory listing of available LLM.txt files with 24-hour local cache
  • OS-specific cache locations:
    • Windows: %LOCALAPPDATA%\llm-txt-mcp
    • macOS: ~/Library/Caches/llm-txt-mcp
    • Linux: ~/.cache/llm-txt-mcp
  • Tools/APIs:
    • list_llm_txt: returns numeric-ID-indexed list of available llm.txt files
    • get_llm_txt: fetch content by numeric ID
    • search_llm_txt: multi-query (array of substrings) search with configurable context lines
  • Multi-query, case-insensitive search that returns matched lines, snippets, and which queries matched
  • Integrates with Model Context Protocol / Claude Desktop workflows
  • MIT license

Use cases

  1. Provide contextual file and documentation access to AI assistants so they can reason about repository structure, dependencies, and documentation.
  2. Enable automated tools to index and search LLM.txt manifests across projects or services.
  3. Integrate into developer environments (Claude Desktop/MCP clients) to let models fetch authoritative docs rather than hallucinate.
  4. Bulk search for configuration, API docs, or dependency mentions across many llm.txt entries.

FAQ

  • Why numeric IDs instead of string identifiers?

The implementation uses opaque numeric IDs because language models were observed to hallucinate plausible-looking string IDs (like hostnames). Numeric IDs encourage clients/models to list and confirm available files instead of guessing.

  • Is this project maintained?

No — it is archived. The mcp-get community server registry is no longer actively maintained.

  • How long is data cached locally?

The directory listing is cached for 24 hours by default. Cache locations vary by OS (see Key features).

  • How do I search within an llm.txt file?

Use the search_llm_txt tool with the numeric file ID, an array of query substrings (case-insensitive), and optional context_lines to receive matches, snippets, and matched queries.

  • License?

MIT.