Tavily MCP Server
Real-time web search, data extraction, and site crawling for your AI assistant.
Data last scanned today · Reviewed yesterday
Overview
The Tavily MCP server equips AI agents with robust web intelligence capabilities, including targeted search, content extraction, and structured website mapping. These tools allow models to pull live data, crawl web pages, and generate site maps directly within their workflows, supporting more informed and context-aware responses.
Our verdict
This server carries a high-risk security score of 35/100, primarily due to the lack of a standardized authentication method, meaning anyone with the configuration can potentially access it. Given this security profile, it is best suited for experienced developers who can manage API access securely rather than users requiring a strictly audited environment.
- Supports real-time web search and structured site mapping
- Offers remote server option for easier deployment
- Maintained with 2229 GitHub stars
- Small footprint with only 4 direct dependencies
- High security risk due to missing authentication method
- Community-maintained rather than official vendor software
- Not yet editorially reviewed
Setup
{
"mcpServers": {
"tavily-mcp": {
"args": [
"-y",
"tavily-mcp"
],
"command": "npx"
}
}
}Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.
- No read-only mode — running this grants full read/write access.
- No built-in authentication — anyone with access to this config can invoke every tool it exposes.
Tools
| Tool | Description | Risk |
|---|---|---|
| tavily-search | Real-time web search capabilities. | low |
| tavily-extract | Intelligent data extraction from web pages. | low |
| tavily-map | Powerful web mapping tool that creates a structured map of website. | low |
| tavily-crawl | Web crawler that systematically explores websites. | low |
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf | ||||
| ChatGPT |
Frequently asked questions
›What tools does the Tavily MCP server provide?
It provides tools for real-time web search, intelligent data extraction from web pages, structured web mapping, and systematic web crawling.
›How can I authenticate with the remote MCP server?
You can authenticate by including your Tavily API key in the connection URL, via an Authorization header, or by using the supported OAuth authentication flow.
›How do I set default parameters for my requests?
You can specify default parameters by including a DEFAULT_PARAMETERS header containing a JSON object with your preferred settings, such as max_results or search_depth.
›How do I clear stored OAuth credentials if I need to reauthenticate?
You can clear your stored credentials by running the command 'rm -rf ~/.mcp-auth' in your terminal.
›Can I control which API key is used during the OAuth flow?
Yes, you can designate a specific key to be used by naming it 'mcp_auth_default' within your personal or team account settings.
›Is it possible to make the Tavily MCP server available globally in Claude Code?
Yes, you can add the '--scope user' flag to your 'claude mcp add' command to make the server accessible across all your projects.
Alternatives
Badge
Maintain this server? Add the live badge to your README.