MCP Tavily

🚀 Tav Semantic Context Protocol Server

Tav is a server designed to implement the Model Context Protocol. This protocol defines how to interact with external knowledge sources and the network through language models. The current implementation version is v0.5.0, which supports the following features:

• Search for information from external knowledge bases.
• Support domain filtering via URLs.
• Provide summaries of web page content.
• Generate reports in citation formats (e.g., MLA, APA).
• Directly return results based on web searches.

🚀 Quick Start

📦 Installation

Clone the repository

git clone git@github.com:modelcontextprotocol/mcp-tavily.git
cd mcp-tavily

Run the server

python -m mcp_server_tavily

By default, the server runs on port 8000 and binds to all network interfaces. You can specify a different port using the following command:

TAV_PORT=8080 python -m mcp_server_tavily

Configure environment variables

The following are common environment variable configurations:

• TAV_API_KEY: API key for authentication (required).
• TAV_HOST: The host address to which the server binds, defaulting to 0.0.0.0.
• TAV_PORT: The port on which the server runs, defaulting to 8000.
• OPENAI_API_KEY: OpenAI API key (optional, used for calling GPT models).

✨ Features

1. Knowledge base integration

Supports multiple knowledge base sources:

# Configure Wikipedia search
TAV_KNOWLEDGE_SOURCES="wikipedia"

# Enable network search functionality
TAV_NETWORK=True

2. Domain filtering

Filter content in specified domains via URLs:

curl -X POST http://localhost:8000/api/v1/chat/completions \
-d '{"model": "tav", "messages": [{"role": "user", "content": "Tell me about quantum computing. Please exclude arXiv sources."}]}'

3. Report generation

Supports multiple citation formats:

curl -X POST http://localhost:8000/api/v1/chat/completions \
-d '{"model": "tav", "messages": [{"role": "user", "content": "Write a report on AI ethics in markdown with MLA citations. Include arXiv sources only."}]}'

💻 Usage Examples

Generate a summary

curl -X POST http://localhost:8000/api/v1/chat/completions \
-d '{"model": "tav", "messages": [{"role": "user", "content": "Summarize the latest advancements in quantum computing."}]}'

Return search results

curl -X POST http://localhost:8000/api/v1/chat/completions \
-d '{"model": "tav", "messages": [{"role": "user", "content": "Search for recent breakthroughs in AI and return direct results."}]}'

🧪 Test Suite

The project provides a complete test suite to verify the correctness of functions:

uv sync --dev

After running the tests, you will see output similar to the following:

✓ Basic functionality test - 100ms
✓ Knowledge source integration - 200ms
✓ Citation formatting - 300ms

🐳 Docker Deployment

Build the image

uv package --format docker:image --name tav-semantic-server .

Run the container

docker run -it --rm -p 8000:8000 \
-e TAV_API_KEY=your_api_key \
tav-semantic-server

🛠️ Debugging Tools

The project provides a debugging interface for developers to facilitate debugging:

uv debug .

Access http://localhost:5173 to view real-time logs and debugging information.

🤝 Contribution Guidelines

Contributions are welcome! Please follow these steps:

1. Create a feature branch: git checkout -b feature/your-feature.
2. Commit your changes: git add . && git commit -m "feat: your new feature".
3. Push the branch: git push origin feature/your-feature.
4. Create a Pull Request.

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.

📞 Contact Us

If you have any questions or suggestions, please contact us:

• Email: contact@modelcontextprotocol.org
• GitHub repository: https://github.com/modelcontextprotocol/mcp-tavly