MCP.so
Sign In

tavily-search MCP server

@Tomatio13

About tavily-search MCP server

Taivily APIを利用するMCP Serverです。Officialが提供される前に開発しました

Basic information

Category

Search

License

MIT

Runtime

python

Transports

stdio

Publisher

Tomatio13

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "mcp-server-tavily": {
      "command": "npx",
      "args": [
        "-y",
        "@smithery/cli",
        "install",
        "tavily-search",
        "--client",
        "claude"
      ]
    }
  }
}

Tools

1

Performs searches based on specified queries

Overview

What is tavily-search?

tavily-search is an MCP server that integrates the Tavily API, enabling AI assistants such as Claude Desktop and Cursor to perform web searches. It returns search results as text, including AI-generated answers, URIs, and titles of the results.

How to use tavily-search?

Install by cloning the repository, configuring your MCP client (e.g., Claude Desktop) with your TAVILY_API_KEY in the environment variables, and restarting the client. The server exposes a single tool named search, which requires a query argument and accepts an optional search_depth (basic or advanced). You can also install via Smithery or run in a local Docker Compose environment.

Key features of tavily-search

  • Performs web searches using the Tavily API.
  • Returns AI-generated answers alongside search results.
  • Supports basic and advanced search depth.
  • Integrates with Claude Desktop, Cursor, and MCP clients.
  • Stores logs locally for debugging.
  • Can be run via Docker Compose on any OS.

Use cases of tavily-search

  • Ask an assistant to find today’s events in a specific city.
  • Retrieve up-to-date news or facts for AI-powered conversations.
  • Perform detailed research by specifying advanced search depth.
  • Enable agentic workflows that require live web data.

FAQ from tavily-search

What environment variable is required?

You must set TAVILY_API_KEY to your Tavily API key. Optionally set PYTHONIOENCODING to utf-8.

Where are logs stored?

On Windows, logs are stored in C:\Users\[username]\AppData\Roaming\Claude\logs\mcp-server-tavily-search. The location for other operating systems is not specified in the README.

Can I run tavily-search on operating systems other than Windows or macOS?

Yes, you can run it using Docker Compose. The README provides instructions for setting up a local client in a Docker container.

What arguments does the search tool require?

The search tool requires a query (string) and accepts an optional search_depth (string with values basic or advanced).

Does tavily-search work with Cursor?

Yes, you can create a shell script that sets the environment variables and runs the server via uv, then configure Cursor’s MCP server settings to execute that script.

Comments

More Search MCP servers