llm-mcp

LLM plugin for Model Context Protocol (MCP) support, enabling LLMs to use tools from any MCP server.

Requirements

• LLM version 0.26 or higher (for tool support)
• Python 3.10+

Installation

First, install or upgrade LLM to version 0.26+:

uv tool install llm
# or upgrade if you have it already
uv tool upgrade llm

Then install the plugin:

llm install llm-mcp

Basic Usage

Adding MCP Servers

Add a remote MCP server:

llm mcp servers add "https://gitmcp.io/simonw/llm"
✔ added server 'gitmcp_llm' with 4 tools

Add a local MCP server via npx:

llm mcp servers add "npx @wonderwhy-er/desktop-commander"
✔ added server 'desktop_commander' with 18 tools

Managing Servers

List servers:

llm mcp servers list

View server details:

llm mcp servers view gitmcp_llm

Remove a server:

llm mcp servers remove gitmcp_llm

Using Tools

Once a server is added, its tools become available to use with any LLM model:

# Use a single tool
llm -T read_file "What is the secret word in secret.txt?"

# Use multiple tools
llm -T search_llm_documentation -T fetch_generic_url_content \
  "Find docs for how to specify a schema in llm project"

# Debug mode to see tool calls
llm -T tool_name "your prompt" --td

Roadmap to v0.1

• ✅ v0.0.2 - Basic MCP server management and tool usage
  • `llm mcp servers` for add, list, view
  • Convert MCP tools to LLM tools
  • Support for stdio and HTTP servers

• 🚧 v0.1.0 - Advanced features
  • Remote server authentication (tokens, OAuth)
  • `llm mcp toolboxes` - create and manage tool collections
  • Support vanilla Python functions as tools
  • `llm mcp proxy` - start MCP proxy server for toolboxes
  • Proxy authentication

Resources

• Simon Willison's blog post on LLM 0.26 tool support
• LLM Tools documentation
• Model Context Protocol

Safety Warning

⚠️ Tools can be dangerous! Be careful about which tools you enable when working with untrusted content. See the LLM tools documentation for important security considerations.