Skip to content
PhenoMePhenoMe

MCP for PhenoMe

The MCP server is optional. It does not change PhenoMe defaults and does not run embedding extraction. Install it only if you want Cursor, Claude Desktop, or similar tools to call PhenoMe helpers.

Install

Section titled “Install”
Terminal window
pip install "phenome[mcp]"

From a clone (editable):

Terminal window
pip install -e ".[mcp]"

Then run:

Terminal window
phenome-mcp

Cursor example

Section titled “Cursor example”

In MCP settings, point stdio at the phenome-mcp executable from the same environment where PhenoMe is installed:

{
  "mcpServers": {
    "phenome": {
      "command": "phenome-mcp"
    }
  }
}

Tools

Section titled “Tools”
  • search_docs — keyword search over local docs (if you have a clone), bundled recipes, or the published llms-small.txt
  • get_recipe — named snippets (first-analysis, export-settings, path-metadata, property-presets, batch-correction, checkpoint)
  • list_public_api — top-level imports and PhenoMe methods
  • inspect_dataset — find_files + inspect_data on a folder (no GPU / no embeddings)
  • format_run_settings — pretty-print a JSON file from export_experiment_config

Assistants should still only pass extra PhenoMe parameters (seed, include_run_settings, log level, plot kwargs) when the user asked. See Using with LLMs.