Skip to content

Releases: BaranziniLab/MedCP

MedCP

20 Jan 08:14

Choose a tag to compare

MedCP v1.0.0 Release Notes

Major Update: Standalone Python Runtime & MCPB Format

This release marks a significant architectural improvement to MedCP, transitioning from UV package manager to a fully self-contained standalone Python runtime bundled directly in the extension.

What's New

Standalone Python Runtime

  • No external dependencies required - Python 3.12 runtime and all dependencies are now bundled within the extension
  • Simplified installation - No need to install UV, Homebrew, or manage Python versions
  • Faster startup - Direct Python execution without UV overhead
  • Cross-platform ready - Works on macOS, Windows, and Linux out of the box

MCPB Format Migration

  • Updated to MCPB 0.2 specification for better compatibility
  • Enhanced metadata with comprehensive descriptions
  • Improved discoverability with updated keywords and platform specifications

Bundle Optimizations

  • Optimized package size: 58.8 MB compressed / 156.1 MB unpacked
  • 10,475 files including complete Python runtime and dependencies
  • Proper ignore configuration using .mcpbignore for cleaner packaging

What's Included

  • Python 3.12.8 standalone runtime
  • FastMCP 2.14.3 - Latest MCP framework
  • Neo4j 6.1.0 - SPOKE knowledge graph connectivity
  • pymssql 2.3.11 - SQL Server/OMOP CDM integration
  • pydantic 2.12.5 - Data validation
  • 90+ additional dependencies - All fully bundled and ready to use

Technical Improvements

Manifest Updates

  • Changed from dxt_version to manifest_version: "0.2"
  • Added long_description field with comprehensive extension overview
  • Explicit Python 3.12 specification in command path
  • Added compatibility section with platform and runtime requirements
  • Enhanced keywords for better discoverability

Code Fixes

  • Fixed FastMCP API compatibility (removed deprecated dependencies parameter)
  • Updated Python command to explicitly use python3.12

Documentation Updates

  • Updated all references from DXT to MCPB format
  • Removed UV installation instructions (no longer needed)
  • Simplified setup guide - just download Claude Desktop and install the extension
  • Updated troubleshooting guide for MCPB extensions

Installation

  1. Download MedCP.mcpb from this release
  2. Double-click the file to install in Claude Desktop
  3. Configure your database credentials in the setup wizard
  4. Start using MedCP immediately

Configuration

MedCP supports two data sources (at least one required):

SPOKE Knowledge Graph (Neo4j)

  • Drug-disease associations
  • Protein interactions
  • Biological pathways

Clinical Records (SQL Server/OMOP CDM)

  • Electronic health records
  • Patient demographics
  • Clinical outcomes

All credentials are encrypted and stored securely in your OS keychain.

Security

  • Local processing only - All data stays on your machine
  • Encrypted credentials - OS keychain integration
  • Read-only database access - Safe querying without modification risks

Bundle Details

  • Package size: 58.8 MB (compressed)
  • Unpacked size: 156.1 MB
  • Total files: 10,475
  • SHA256: 217fa913817139443dae888dd791673eac723dae

Acknowledgments

Built by the Baranzini Lab at UCSF

  • SPOKE Knowledge Graph: Morris et al., 2023
  • UCSF Clinical Data: UCSF Research Data Portal

License

MIT License

MedCP

03 Oct 21:21

Choose a tag to compare

Fixed minor bugs with the updated version of Anthropic Claude

MedCP

01 Oct 00:33

Choose a tag to compare

Minor bug fix: using native Cypher queries to extract knowledge graph schema

MedCP

09 Aug 22:53

Choose a tag to compare

Fixing database compatibility issues