Official Tavily plugin for Cursor. Adds web search, content extraction, website crawling, and AI-powered research capabilities.
| Skill | Command | Description |
|---|---|---|
| Search | /search |
Search the web using Tavily's LLM-optimized API. Returns relevant results with content snippets, scores, and metadata. |
| Research | /research |
Comprehensive research on any topic with citations. Supports mini and pro models. |
| Extract | /extract |
Extract clean content from specific URLs. Returns markdown/text from web pages. |
| Crawl | /crawl |
Crawl websites to download documentation, knowledge bases, or web content as local markdown files. |
| Best Practices | /tavily-best-practices |
Reference documentation for building production-ready Tavily integrations. |
No manual setup required — Uses OAuth via the Tavily MCP server.
You must have an existing Tavily account. The OAuth flow only supports login — account creation is not available through this flow. Sign up at tavily.com first if you don't have an account.
On first run, the script will:
- Check for existing tokens in
~/.mcp-auth/ - If none found, automatically open your browser for OAuth authentication
Get an API key at https://tavily.com and set it as an environment variable:
TAVILY_API_KEY=tvly-your-api-key-here
MIT