llm-tools-mcp

MCP tools support

Usage

To use this with the LLM command-line tool:

$ echo "The magic word is frumble" > magic.txt
$ uv run llm --tool 'MCP(command="npx", args=["-y", "@modelcontextprotocol/server-filesystem", "."])' "What is the magic word in the file ./magic.txt?" --tools-debug
Secure MCP Filesystem Server running on stdio
Allowed directories: [ '/Users/aw/Projects/rectalogic/llm-tools-mcp' ]

Tool call: read_file({'path': './magic.txt'})
  "CallToolResult(meta=None, content=[TextContent(type='text', text='The magic word is frumble\\n', annotations=None)], isError=False)"

The magic word in the file `magic.txt` is **frumble**.

With the LLM Python API:

import llm
from llm_tools_mcp import MCP

model = llm.get_model("gpt-4.1-mini")

result = model.chain(
    "Example prompt goes here",
    tools=[MCP(command="npx", args=["-y", "@modelcontextprotocol/server-filesystem", "."])]
).text()

Development

To set up this plugin locally, first checkout the code. Then create a new virtual environment:

cd llm-tools-mcp
uv sync

To run the tests:

uv run pytest