FastMCP Tool Server Starter

Tiny FastMCP server that imports public functions from a science package and exposes them as MCP tools.

Requires Python 3.12+.

Install

python3.12 -m venv .venv
source .venv/bin/activate
pip install -e ./science-demo
pip install -e .

Or with uv that is faster and tidy than pip, but optional:

uv sync

Run

For agents that launch the server as a subprocess:

fast-mcp-starter --transport stdio

For agents that connect to a running MCP URL:

fast-mcp-starter --transport streamable-http --host 127.0.0.1 --port 8000

Endpoint:

http://127.0.0.1:8000/mcp

Test With A LangGraph Agent

See notebooks/test_with_agent.ipynb for a small external LangGraph agent that connects to this MCP server.

Hook In Your Repo

Expose normal public Python functions in your external package, add that package as a dependency in pyproject.toml, then import and register those functions in server.py and server.py.

Keep science-demo as a working example while you add your own repo. Delete it later only if you do not want the demo tools.

This starter does that with:

from science_demo.science_tools import generate_random_points, plot_sine_wave

mcp.tool()(generate_random_points)
mcp.tool()(plot_sine_wave)

Demo Tools

The local science-demo package exposes:

generate_random_points(output_dir: str, count: int = 30, seed: int = 42) -> ArtifactResult
plot_sine_wave(output_dir: str, num_points: int = 200) -> ArtifactResult

Both return typed structured content:

{"status": "success", "files": ["..."], "message": "...", "metadata": {}}

Test

python -m pytest