Skip to content

implementation_plan.md

Latest commit

 

History

History
50 lines (30 loc) · 2.44 KB

implementation_plan.md

File metadata and controls

50 lines (30 loc) · 2.44 KB
Raw

Implementation Plan - Bio-MCP ClinVar Server

Goal Description

Build a Bio-MCP (Model Context Protocol) Server that functions as an intelligent genomic agent tool. This server will expose tools to query the ClinVar database (via NCBI E-utilities) to retrieve and interpret variant clinical significance. This serves as a portfolio artifact (Bio-Portfolio) demonstrating "Bio-AI" capabilities and Agentic integration.

User Review Required

[!IMPORTANT] > Tech Stack Selection: We are using Python (with uv for management) as it is the standard for Bioinformatics/Data Science. Dependency: External access to NCBI E-utilities API (public). Supabase Alignment: This MCP server is designed to be potentially registered in a Supabase-backed Tool Registry, adhering to the "Reasoning Auditor" persona by providing structured, stateless tool execution.

Proposed Changes

Project Structure

New directory: Bio-MCP-ClinVar/

[NEW] pyproject.toml

  • Define project dependencies: mcp, httpx, xmltodict (for E-utilities parsing).

[NEW] src/main.py

  • MCP Entry Point: Implements the MCP server using FastMCP or standard mcp server classes.
  • Tools:
    • search_clinvar(term: str): Searches ClinVar for a term (gene, variant).
    • get_variant_details(uid: str): Retrieves clinical significance and summary for a specific variant ID.

[NEW] src/clinvar.py

  • API Client: Handles HTTP requests to https://eutils.ncbi.nlm.nih.gov/entrez/eutils/.
  • Parsing: Converts XML responses to JSON/Dict for the agent.

[NEW] README.md

  • Documentation: Instructions on how to run the server locally and use it with an MCP client (e.g., Claude Desktop or generic MCP client).

Verification Plan

Automated Tests

  • Create a simple test script test_server.py to verify API connectivity and response parsing.
  • Run uv run pytest (if full tests) or a manual python script.

Manual Verification

  1. Install: Run uv sync.
  2. Run: Start server uv run python src/main.py.
  3. Inspect: Use an MCP inspector or dry-run script to call search_clinvar and verify output.