home / mcp / mcp tavily mcp server

MCP Tavily MCP Server

mcp server of tavily

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "kshern-mcp-tavily": {
      "command": "npx",
      "args": [
        "-y",
        "@mcptools/mcp-tavily"
      ],
      "env": {
        "TAVILY_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

You can run a Tavily-enabled MCP server that provides powerful search capabilities and content extraction through a simple, configurable MCP interface. This server lets you perform basic, context-aware, and Q&A searches, and extract content from URLs with customizable depth and options, all accessible via MCP client calls.

How to use

Set up the Tavily MCP server in your MCP configuration, then call its tools from your MCP client by name. You can perform a Basic Search, a Context Search, or a Q&A Search, each with adjustable options like search depth and topics. You can also request content extraction from URLs with configurable depth and image inclusion. Use the Tavily tools to build complex search workflows, combine results, and pull relevant content for your applications or research.

How to install

Prerequisites: you need Node.js 16 or higher and npm or yarn. You also require a Tavily API key to access Tavily services.

# Install the Tavily MCP server locally
npm install @mcptools/mcp-tavily

# Or run it directly with npx
npx @mcptools/mcp-tavily

# Set your Tavily API key (example)
export TAVILY_API_KEY=your_api_key

Additional setup and run notes

Build and start flow is described in the typical usage pattern where you install the package, configure the environment, and then start the server. A direct start example shown is to run the MCP Tavily package via npx, which uses your API key to authenticate with Tavily.

Configuration and usage notes

Configure the Tavily MCP server by including it in your MCP configuration with the following command elements: the command is npx and the arguments include the Tavily MCP package, along with your API key as an environment variable.

Testing and debugging with MCP Inspector

For development and debugging, you can use MCP Inspector. It provides an interface to test tool calls, view server responses, and monitor the server state during development.

Available tools

search

Basic search tool that returns results for a given query with configurable options like depth, topic, and max results.

searchContext

Context-aware search tool that adjusts relevance based on topic and time range to improve result quality.

searchQNA

Question and answer style search tool that retrieves answers with optional inclusion of the answer content and a limited number of results.

extract

Content extraction tool that retrieves content from one or more URLs with options for extraction depth and image inclusion.