Proteomics MCP

Proteomics MCP is a local-first MCP server and Python package for inspecting proteomics identification results and selected MS2 spectra. It normalizes outputs from InstaNovo, Casanovo, MaxQuant, FragPipe, Proteome Discoverer, and user-mapped tables, then supports targeted spectrum retrieval, fragment prediction, scoring, annotation, and visualization.

It is an inspection and evidence layer, not a de novo sequencing model, search engine, FDR workflow, or protein-inference pipeline.

Install

Python 3.11 to 3.14 is supported. Install from PyPI:

pip install proteomics-mcp
proteomics-mcp --help

To run the server from an MCP-capable client without installing it in the current environment:

uvx proteomics-mcp serve

Platform requirements (glibc for Python 3.13+ on Linux, macOS versions, and brew install libomp for local MS2PIP on macOS) are listed in the README.

The server uses stdio and must be launched by an MCP client, not run interactively in a terminal. For Claude Desktop, the GitHub release also contains a .mcpb extension.

Typical workflow

  1. Inspect an upstream PSM table and its available fields.
  2. Select a peptide, PSM, or constrained subset.
  3. Retrieve the corresponding MS2 spectrum, predict fragments if needed, and score or annotate the candidate.
  4. Render a spectrum or aggregate saved scores into a table or dashboard.

Start with the technical reference for formats and commands, or the prompt examples for agent workflows.

Data and privacy

Local input and output paths remain on your computer unless you explicitly use a network-backed tool. Koina predictions, USI retrieval, and PRIDE metadata lookup contact external services. Interactive HTML reports load Plotly from a CDN when opened. Review the privacy policy before using sensitive data.

Example outputs are available in the source repository. The manuscript's supplementary data and figure evidence include processed identification and prediction tables, search settings, rendered figures, and an inventory of exact raw-file names. Original vendor RAW files are available from PRIDE under PXD044934, PXD014877, and PXD010595. Raw datasets and large converted spectra are not distributed with the Python package.