PhytoData-MCP

Standalone MCP Server for Plant Genomics

PhytoData-MCP solves the problem of siloed plant genomics databases by exposing the four most important plant genomics databases as Model Context Protocol (MCP) tools. This allows LLMs (like Claude) to autonomously chain queries together and build a complete picture of a single gene's function, pathway membership, cross-species conservation, and protein behavior.

Features

• PhytoMine (Phytozome): Query GO terms, Pfam domains, and KO IDs for given genes.
• KEGG: Look up pathways from KO IDs or EC numbers.
• Ensembl Plants: Find orthologous genes across multiple crop species.
• UniProt (Upcoming): Subcellular localization and protein domains.

Installation

You can install phytodata-mcp directly from python:

pip install phytodata-mcp

Development Installation

To install for development, with testing dependencies:

git clone https://github.com/zaeyasa/phytodata-mcp.git
cd phytodata-mcp
pip install -e .[dev]

Claude Desktop Configuration

Add the server to your claude_desktop_config.json:

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

Usage Example

Once connected, you can ask Claude questions like:

"Fetch the gene info for AT3G24650 in Arabidopsis, identify its KEGG pathway, and find its orthologs in wheat and tomato."

The LLM will automatically chain:

1. phytomine_gene_info -> Retrieves KO ID
2. kegg_pathway_lookup -> Uses KO ID to find pathways
3. ensembl_plants_orthologs -> Matches gene across target crops

Tools Reference

• phytomine_gene_info: Requires gene_id and organism.
• kegg_pathway_lookup: Requires ko_id or ec_number.
• ensembl_plants_orthologs: Requires gene_id and target_species (can use "all_crops").

License

This project is licensed under the MIT License.