Tavily

You can run an MCP server that lets you perform AI-powered web searches through Tavily’s API, with proxy support. This server enables LLMs to fetch fresh results via proxies, extract relevant content, and generate direct answers or news summaries for your applications.

How to use

Use an MCP-enabled client to connect to the Tavily MCP Server and issue search or news queries. You can perform full web searches with AI-extracted content, generate concise answers with supporting evidence, or pull recent news items. The server handles proxy configuration automatically if you set the appropriate environment variables, so your queries reach Tavily through your proxy setup. Use simple, natural language prompts to tell the server what you want searched or summarized.

How to install

Prerequisites: You need Python installed on your system. You also need a Tavily API key to access Tavily’s services.

Install the Tavily MCP Proxy package using Python’s package manager.

pip install mcp-tavily-proxy

If you have the UV tool installed, you can use it to run the installer via UV.

uv pip install mcp-tavily-proxy

Build from source if you prefer to compile locally.

git clone https://github.com/tulong66/mcp-tavily-proxy.git
cd mcp-tavily-proxy
uv build
uv pip install .

If you want to run directly from the source using Python, you can start the server with the API key supplied on the command line.

python -m mcp_server_tavily --api-key=your_api_key_here

Configure environment variables for proxy access and API key before starting the server. You can export them in your shell session.

export TAVILY_API_KEY=your_api_key_here
export HTTP_PROXY=http://your-proxy:port
export HTTPS_PROXY=http://your-proxy:port

Or inject the API key via a CLI argument at startup.

python -m mcp_server_tavily --api-key=your_api_key_here

Add the Tavily MCP server to your Claude.app configuration to wire it into your workflows.

{
  "mcpServers": {
    "tavily": {
      "command": "python",
      "args": ["-m", "mcp_server_tavily"]
    },
    "env": {
      "TAVILY_API_KEY": "your_api_key_here",
      "HTTP_PROXY": "http://your-proxy:port",
      "HTTPS_PROXY": "http://your-proxy:port"
    }
  }
}

If you need to debug proxy configurations, enable debug logging and start the server to see proxy details.

export TAVILY_LOG_LEVEL=DEBUG
python -m mcp_server_tavily

You can also inspect the MCP setup with the MCP inspector tool.

npx @modelcontextprotocol/inspector python -m mcp_server_tavily

Configuration notes and examples

The server uses an API key for Tavily and supports common HTTP proxies. If you need to run behind a proxy, set HTTP_PROXY and HTTPS_PROXY in your environment or pass them via the environment field shown in the configuration example.

Examples of common use cases include performing a regular web search, generating a detailed report with explicit formatting and citations, or retrieving the latest AI-assisted news within a defined window of days.

Available tools

tavily_web_search

Performs comprehensive web searches with AI-powered content extraction. Parameters: query, max_results, search_depth.

tavily_answer_search

Performs web searches and generates direct answers with supporting evidence. Parameters: query, max_results, search_depth.

tavily_news_search

Searches recent news articles with publication dates. Parameters: query, max_results, days.