Open Targets MCP Server 🧬

A professional-grade, workflow-oriented Model Context Protocol (MCP) server for therapeutic discovery and drug repurposing. This server provides a seamless bridge between agentic AI platforms and the Open Targets Platform, leveraging its comprehensive GraphQL API.

---

🚀 Key Features

• Modular Architecture: Built with a clean separation of concerns (Clients, Tools, Models) following the Top 1% MCP standard.
• Type-Safe Validation: Powered by Pydantic to ensure 100% reliable data structures for AI consumption.
• Workflow-Oriented Tools: Abstracted complex queries into high-level therapeutic workflows (Cancer profiling, Safety, Tractability).
• Modern Transport: Supports the latest Streamable-HTTP transport layer.
• Automated Testing: Integrated pytest suite for verified API reliability.

---

🏗 Repository Structure

├── src/
│   └── opentarget_mcp/
│       ├── clients/       # GraphQL & API Communication
│       ├── tools/         # Domain-specific logic (Analysis, Cancer, etc.)
│       ├── models/        # Pydantic data validation classes
│       └── server.py      # Modular server initialization
├── tests/                 # Automated test suite (pytest)
├── server.py              # Root entry point
└── requirements.txt       # Optimized dependencies

---

🛠 Installation & Setup

1. Clone & Environment

git clone https://github.com/aakashyadav27/OpentargetMCP.git
cd OpentargetMCP
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Run the Server

export PYTHONPATH=$PYTHONPATH:.
python3 server.py

The server will start using the Streamable-HTTP transport, listening at http://127.0.0.1:8000/mcp.

---

🧪 Testing

This project includes a comprehensive test suite to ensure API compliance.

export PYTHONPATH=$PYTHONPATH:.
pytest tests/

---

🤖 Usage with AI Agents

Option A: Automatic Setup (Recommended)

This method tells the AI client (Claude Desktop, Cursor, etc.) to start the server for you in the background.

{
  "mcpServers": {
    "opentarget": {
      "command": "python3",
      "args": ["/absolute/path/to/OpentargetMCP/server.py"],
      "env": {
        "PYTHONPATH": "/absolute/path/to/OpentargetMCP"
      }
    }
  }
}

Option B: Remote / Manual Setup

Use this if you are running the server manually in a terminal or on a remote machine.

1. Start the server:

   export PYTHONPATH=.
   python3 server.py

2. Add this to your config:

   {
     "mcpServers": {
       "opentarget": {
         "url": "http://127.0.0.1:8000/mcp"
       }
     }
   }

---

📄 License

MIT License