🧠 ToolUniverse MCP Server — Dockerized Setup

Run the ToolUniverse MCP Server from Harvard Zitnik Lab inside a lightweight Docker container.
ToolUniverse is an AI-powered biomedical multi-agent system integrating public datasets, APIs, and tools for life science research.

This image lets you deploy tooluniverse.smcp_server (the official MCP server) quickly and securely in any environment — local, cloud, or HPC cluster.

---

🚀 Features

• ✅ Ready-to-run MCP (Model Context Protocol) server for ToolUniverse
• 🧩 Based on official Python 3.10 slim image
• 🔐 Supports API keys for OpenTargets, NCBI, USPTO, NVIDIA, Hugging Face, etc.
• 🧬 Designed for AI biomedical research, data integration, and agent collaboration
• ⚙️ Compatible with Docker Compose & .env secrets
• 🧱 Minimal footprint (~200MB)

---

📦 Installation

1. Clone or copy this repository

git clone https://github.com/yourusername/tooluniverse-mcp-docker.git
cd tooluniverse-mcp-docker

2. (Optional) Create .env file

Create .env file to securely store your API keys:

OPENTARGETS_API_KEY=your_key
NCBI_API_KEY=your_key
USPTO_API_KEY=your_key
HF_TOKEN=your_token
NVIDIA_API_KEY=your_key

3. Build the Docker image

docker build -t tooluniverse-mcp .

4. Run the container

docker run -p 8000:8000 --env-file .env tooluniverse-mcp

or via Docker Compose:

docker compose up -d

---

🐳 Dockerfile (Minimal Example)

FROM python:3.10-slim

WORKDIR /app

RUN pip install --no-cache-dir tooluniverse

EXPOSE 8000

CMD ["python", "-m", "tooluniverse.smcp_server"]

---

🧰 Configuration (Environment Variables)

ToolUniverse MCP uses various APIs for data retrieval and model access.
You can provide them either via .env or Docker Compose:

Variable	Description
OPENTARGETS_API_KEY	Access to OpenTargets biomedical data
NCBI_API_KEY	Access to NCBI databases
USPTO_API_KEY	Access to USPTO patent data
HF_TOKEN	Hugging Face model hub authentication
NVIDIA_API_KEY	Access to NVIDIA BioNeMo services
USPTO_MCP_SERVER_HOST	Optional external MCP host for USPTO
TXAGENT_MCP_SERVER_HOST	Optional host for TXAgent
EXPERT_FEEDBACK_MCP_SERVER_URL	Local feedback service URL

---

🌐 Access

Once the container is running, your MCP server will be available at:

http://localhost:8000

Use it directly or connect it to other MCP-compatible clients (e.g. AI agents, OpenAI tools, or research assistants).

---

🔍 SEO Keywords

ToolUniverse MCP, ToolUniverse Docker, MCP server setup, Harvard Zitnik Lab, biomedical AI, multi-agent system, Model Context Protocol, AI research tools, biological data integration, Docker MCP server

This project makes it easy to deploy ToolUniverse MCP for AI-driven biomedical research, allowing integration with multi-agent frameworks and data APIs in a reproducible Docker environment.

---

🧪 About ToolUniverse

ToolUniverse is developed by the Zitnik Lab at Harvard Medical School.
It connects hundreds of biomedical tools, models, and datasets into one unified multi-agent platform for discovery and hypothesis generation.

Learn more: https://zitniklab.hms.harvard.edu/ToolUniverse

---

📝 License

MIT License — see LICENSE for details.

---

💡 Citation

If you use ToolUniverse in your research, please cite the original Harvard Zitnik Lab project.

@software{zitniklab_tooluniverse_2024,
  title = {ToolUniverse: Multi-Agent AI System for Biomedical Research},
  author = {Marinka Zitnik Lab, Harvard Medical School},
  year = {2024},
  url = {https://zitniklab.hms.harvard.edu/ToolUniverse}
}

---

🧩 Related Links

• ToolUniverse Documentation
• Harvard Zitnik Lab GitHub
• Model Context Protocol (MCP)
• OpenTargets Platform
• Hugging Face Models

---

❤️ Contributing

Pull requests and issues are welcome!
Feel free to fork this repo, submit PRs, or open feature requests for MCP or Docker improvements.

---

Maintained by: [Your Name or Org]
Based on: Harvard Zitnik Lab — ToolUniverse Project