M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
T

Tavily MCP Server

Official
Production-ready MCP server with real-time search, extract, map & crawl.
How to install ↓ Compare GitHub source repository ↗
Category
Other #88 of 230
Stars
★ 2.4k Very popular
Transport
stdio (local process) · Streamable HTTP
Runtime
Node.js 20+ · Hosted (no local runtime)
Credentials
API key / credential required
License
MIT
Last commit
Tools
4
58FMRS · C

The Tavily MCP server is used to provide real-time web search and extraction to AI assistants, offering local or remote deployment and support for multiple clients.

Strongest · Maintenance 13/20 Weakest · Reliability 8/20

Reliability
8/20
Security and permissions
13/20
Maintenance
13/20
Documentation
12/20
Setup experience
12/20
Why each score
Reliability 8/20
Source review shows TypeScript source and tests, with CI workflow files present. However, the server manifest and exact tool implementations were not fully cross-verified, and the README claims (e.g., crawler) depend on external API key and services. Static review cannot confirm runtime behavior. Happy-path startup and tool list appear plausible, but error handling and edge cases are not thoroughly demonstrated, so reliability is capped below 12.
Security and permissions 13/20
No malicious code or credential theft observed. API keys are passed via environment variables or URL headers, and docs advise against hardcoding. The server makes external network calls to Tavily API, which is its purpose. No irreversible operations or dangerous commands without confirmation. However, user data is sent to external service, and output data boundaries are not clearly defined. Least privilege is not explicitly discussed, but no clear overreach. Security score is moderately high.
Maintenance 13/20
Repository is officially maintained by Tavily, with active commits and releases, and MIT license. However, open issues count is high (50), and no explicit security response channel is documented. Stars and npm downloads indicate community activity, but governance and update path visibility is partial.
Documentation 12/20
README provides detailed installation, authentication (API key or OAuth), remote and local server config, environment variables, example client configs, and tutorial links. However, it lacks thorough troubleshooting and does not mention API rate limits, costs, or error-handling limitations. Docs are good but with room for improvement.
Setup experience 12/20
Simple install via npx or remote URL, with configuration examples for mainstream clients like Claude Code and Cursor. However, requires manual API key acquisition and configuration, and no step-by-step installer or automated test. Setup is clear but manual.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessUses the network

Best for

  • Developers needing real-time web capabilities to enhance AI responses.
  • Users working within Claude Code, Cursor, or any compatible MCP client.
  • Teams that require quick and reliable data extraction and search.

Not for

  • Offline applications that don't require real-time web features.
  • Scenarios requiring extensive crawling that may exceed free tier limits and need careful management.
  • Environments with strict data privacy concerns where sending data to external services is problematic.

Required permissions

  • Access to real-time web search and extraction APIs.
  • Data transmission over the internet.
  • Requires a valid Tavily API key for authentication.

Risks and side effects

  • API keys might be misused or leaked.
  • Potential costs depending on the Tavily plan's quota.
  • May encounter rate limits or service outages.
  • Do not treat this server as fully trustworthy; verify its behavior before use.

Setup

Before you start

Runtime：Node.js 20+ · Hosted (no local runtime)

TAVILY_API_KEY requiredsecret Tavily API key, obtained by signing up for a free account at tavily.com; required for local runs.
Other optional settings (2)
DEFAULT_PARAMETERS optional Optional JSON of default tavily-search parameters such as include_images, max_results, search_depth.
TAVILY_HUMAN_ID optional Optional identifier for the end user enabling per-user analytics; forwarded as an X-Human-Id header and hashed server-side.
  1. Get a Tavily API key from https://app.tavily.com/home.
  2. For local installation, ensure Node.js v20+ and run npx -y tavily-mcp@latest.
  3. For remote installation, use the URL https://mcp.tavily.com/mcp/?tavilyApiKey=<your-api-key>.
  4. Configure your MCP client (e.g., Claude Code, Cursor) with the server.
claude_desktop_config.json 
{
  "mcpServers": {
    "tavily-mcp": {
      "command": "npx",
      "args": ["-y", "tavily-mcp@latest"],
      "env": {
        "TAVILY_API_KEY": "your-api-key-here",
        "DEFAULT_PARAMETERS": "{\"include_images\": true, \"max_results\": 15, \"search_depth\": \"advanced\"}"
      }
    }
  }
}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "tavily-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "tavily-mcp@latest"
      ],
      "env": {
        "TAVILY_API_KEY": "your-api-key-here",
        "DEFAULT_PARAMETERS": "{\"include_images\": true, \"max_results\": 15, \"search_depth\": \"advanced\"}"
      }
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add tavily-mcp -e TAVILY_API_KEY=your-api-key-here -e 'DEFAULT_PARAMETERS={"include_images": true, "max_results": 15, "search_depth": "advanced"}' -- npx -y tavily-mcp@latest

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

After configuring, the client's tool list should show tavily-search, tavily-extract, tavily-map and tavily-crawl; running one search for real-time information confirms the connection works.

Troubleshooting

  1. Check that the API key is correct and valid.
  2. Ensure network connectivity is stable.
  3. Verify environment variables like `TAVILY_API_KEY` and `DEFAULT_PARAMETERS` are set correctly.
  4. If using OAuth, clear `~/.mcp-auth` to re-authenticate.
  5. Ensure Node.js version meets requirements (v20+).

Things to try

Once connected, you can ask your AI assistant things like:

  • Use tavily-search to find the latest news about the MCP protocol, up to 15 results
  • Extract the main content of https://example.com
  • Create a structured map of https://docs.example.com
  • Crawl this blog site and list all article links

Tools 4

tavily-search read-only
Real-time web search tool that provides up-to-date information.
tavily-extract read-only
Intelligently extracts structured data from web pages.
tavily-map read-only
Creates a structured map of a website to understand its layout.
tavily-crawl read-only
Systematically explores websites to collect content.

Use cases

Providing real-time web search to AI agents for current information.
Extracting structured data from specific web pages for knowledge graphs or analysis.
Generating website maps to understand site structure.
Crawling websites to collect comprehensive data.

Supported clients

Claude Desktop
Claude Code
Cursor

Listed from the project's documentation, not tested by this site.

Overview

The Tavily MCP server provides real-time web search, data extraction, website mapping, and crawling tools for AI agents. It can be run locally via npx or accessed through a remote server URL, with support for API key and OAuth authentication.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 654 · Tools 11 Compare with this →
Compare with other MCP servers

Source revision 1c1d54c2a619 Data synced 2026-10-12 Read the FMRS scoring method