MCP 服务器

BioLab MCP Server

Unified MCP server providing AI-agent-ready access to AlphaFold, PubMed, ChEMBL, Ensembl, and 37+ scientific databases.

README

🧬 BioLab MCP Server

The Stripe for Scientific Data — Unified MCP server providing AI-agent-ready access to AlphaFold, PubMed, ChEMBL, Ensembl, and 37+ scientific databases.

What is BioLab MCP?

BioLab MCP is a premium Model Context Protocol server that gives AI agents (Claude, Cursor, etc.) direct access to the world's leading scientific databases through a single, reliable interface.

Problem: Researchers and AI developers need data from 10+ incompatible APIs with different auth methods, rate limits, and response formats.

Solution: One MCP server. One API key. All databases. Validated, structured responses.

🚀 Quick Start

1. Install

git clone https://github.com/your-org/biolab-mcp-server
cd biolab-mcp-server
uv sync

2. Configure Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "biolab": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/biolab-mcp-server", "biolab-server"]
    }
  }
}

3. Ask Claude

"Search AlphaFold for TP53 protein structure and confidence scores"
"Find the top 10 papers on CRISPR cancer therapy from 2023-2024"
"What drugs target EGFR? Show me IC50 values from ChEMBL"
"Are there any Phase 3 trials recruiting for lung cancer with pembrolizumab?"

🔬 Available Tools (8 MCP Tools)

Tool	Description	Databases
`query_alphafold`	Protein structure predictions with pLDDT scores	AlphaFold DB, UniProt
`search_pubmed`	Scientific literature with metadata extraction	PubMed, PMC
`query_chembl`	Drug compounds, IC50/Ki bioactivity data	ChEMBL
`get_genomic_data`	Gene locations and variant frequencies	Ensembl, gnomAD
`search_proteins`	Protein metadata and interaction networks	UniProt, STRING
`get_clinical_trials`	Clinical study search and monitoring	ClinicalTrials.gov
`find_drug_targets`	Disease-gene associations with confidence scores	Open Targets
`find_related_papers`	Cross-database semantic literature search	PubMed + Semantic Scholar + EuropePMC

💰 Pricing

Tier	Price	Requests/Month	Databases
Free	$0/month	100	AlphaFold, PubMed, ChEMBL
Researcher	$24/month	5,000	All 37+
Lab	$99/month	50,000	All 37+ + Team keys
Enterprise	Custom	Unlimited	All + Private deploy

🛠️ Development

Prerequisites

Python 3.11+
uv package manager

Setup

# Clone and install
git clone https://github.com/your-org/biolab-mcp-server
cd biolab-mcp-server
uv sync --dev

# Copy environment config
cp .env.example .env

# Generate an API key
uv run biolab generate-key --tier researcher --label "my-key"

Run Tests

uv run pytest tests/ -v

Project Structure

src/biolab/
├── server.py          # MCP Server (stdio transport, 8 tools)
├── auth.py            # API key management (SQLite)
├── cache.py           # Disk-based caching
├── config.py          # Settings via pydantic-settings
├── models.py          # Pydantic response models
├── cli.py             # CLI tools
└── tools/
    ├── alphafold.py   # AlphaFold Database
    ├── pubmed.py      # PubMed / PMC
    ├── chembl.py      # ChEMBL drug database
    ├── genomics.py    # Ensembl + gnomAD
    ├── proteins.py    # UniProt + STRING
    ├── clinical.py    # ClinicalTrials.gov
    ├── targets.py     # Open Targets
    └── literature.py  # Cross-database search

🏗️ Architecture

Claude / Cursor (MCP Client)
        ↕ JSON-RPC 2.0 (stdio)
BioLab MCP Server (FastAPI + MCP SDK)
        ↕ HTTP + caching
Scientific Databases:
  AlphaFold DB ─ PubMed/PMC ─ ChEMBL
  Ensembl ─ gnomAD ─ UniProt ─ STRING
  ClinicalTrials.gov ─ Open Targets
  Semantic Scholar ─ Europe PMC