agents-libagents-lib
AgentsMCP ServersWorkflowsTutorialsComparisonsTemplatesNewsAbout
🔌

Tavily Search MCP

Search•⭐ 620

AI-native search API optimized for LLM agents with deep search and answer extraction.

GitHubWebsite →
Claude DesktopCursorWindsurf

Overview

AI-native search API optimized for LLM agents with deep search and answer extraction.

Setup

Run with npx:

npx -y @modelcontextprotocol/server-tavily-search

Configuration

TAVILY_API_KEY environment variable

Documentation

Tavily Search MCP

Overview

Tavily Search MCP provides access to Tavily's AI-native search API, specifically designed for LLM agents. Unlike traditional search APIs, Tavily returns structured results with extracted answers, making it ideal for AI agent workflows that need to understand and synthesize search results.

Features

  • AI-Optimized Results: Returns search results with extracted answers and context
  • Deep Search: Performs multi-step research with source verification
  • Answer Extraction: Gets direct answers from search results
  • Source Attribution: Full citation of sources for fact-checking
  • Search Depth Control: Choose between "basic" and "advanced" search modes
  • Result Filtering: Filter by content type, date range, and domain

Installation

npx -y @modelcontextprotocol/server-tavily-search

Configuration

Add to your Claude Desktop config:

{
  "mcpServers": {
    "tavily-search": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-tavily-search"],
      "env": {
        "TAVILY_API_KEY": "your-api-key"
      }
    }
  }
}

Getting API Credentials

  1. Sign up at Tavily
  2. Get your API key from the dashboard
  3. Free tier includes 1000 searches/month

Available Tools

ToolDescription
searchPerform a web search with Tavily's AI-optimized algorithm
deep_researchConduct multi-step research with comprehensive analysis

Usage Examples

Basic Search

Search for: "How does CrewAI work?"

Returns structured results with:

  • Query
  • Answer summary
  • Individual results with titles, URLs, and content snippets
  • Source citations

Deep Research

Deep research on: "Compare LangGraph vs CrewAI for multi-agent systems"

Performs multi-step research:

  1. Initial search to understand the topic
  2. Follow-up searches for specific aspects
  3. Synthesis of findings into comprehensive answer

Search with Filters

Search with:
- Query: "LangChain documentation"
- Search depth: advanced
- Max results: 10
- Include domains: ["python.langchain.com"]

Claude Desktop Setup

  1. Install the MCP server
  2. Add configuration to claude_desktop_config.json
  3. Restart Claude Desktop
  4. The Tavily Search tool will appear in your available tools

Pros

  • ✅ AI-native design optimized for LLM consumption
  • ✅ Built-in answer extraction and synthesis
  • ✅ Deep research capability for complex queries
  • ✅ Excellent source attribution
  • ✅ Generous free tier (1000 searches/month)
  • ✅ Fast response times

Cons

  • ❌ Requires API key (paid for high volume)
  • ❌ Less customizable than raw search APIs
  • ❌ Results may differ from Google directly
  • ❌ Advanced features require paid plan

Resources