Tavily Cursor MCP Server
Enables web search, content extraction, question answering, and RAG context generation using Tavily API with Cursor-compatible underscore-named tools.
README
Tavily Cursor MCP Server
A custom Tavily MCP server with underscore-named tools for Cursor compatibility.
Why This Exists
The official Tavily MCP server uses hyphenated tool names (tavily-search, tavily-extract) which Cursor's CallMcpTool interface doesn't properly recognize. This custom server uses underscore naming (tavily_search, tavily_extract) to work seamlessly with Cursor.
Features
- ✅ tavily_search - Web search with Tavily API
- ✅ tavily_extract - Extract clean content from URLs
- ✅ tavily_search_qna - Direct question answering
- ✅ tavily_search_context - Generate context for RAG applications
Installation
Option 1: Local Installation (Recommended)
-
Clone or download this directory to your local machine
-
Install dependencies:
cd tavily-cursor-mcp npm install -
Make the script executable (Mac/Linux):
chmod +x index.js -
Add to your Cursor
mcp.json:{ "mcpServers": { "tavily_cursor": { "command": "node", "args": ["/absolute/path/to/tavily-cursor-mcp/index.js"], "env": { "TAVILY_API_KEY": "your-tavily-api-key-here" } } } }Important: Replace
/absolute/path/to/tavily-cursor-mcp/with the actual full path to this directory.
Option 2: NPM Global Installation
-
Install globally:
cd tavily-cursor-mcp npm install -g . -
Add to your Cursor
mcp.json:{ "mcpServers": { "tavily_cursor": { "command": "tavily-cursor-mcp", "env": { "TAVILY_API_KEY": "your-tavily-api-key-here" } } } }
Configuration
Cursor MCP Configuration Location
- Windows:
%APPDATA%\Cursor\User\globalStorage\mcp.json - Mac:
~/.cursor/mcp.jsonor workspace.cursor/mcp.json - Linux:
~/.cursor/mcp.jsonor workspace.cursor/mcp.json
Get Your Tavily API Key
- Go to https://tavily.com
- Sign up or log in
- Get your API key from the dashboard
- Replace
your-tavily-api-key-herein the config with your actual key
Usage in Cursor
After installation and configuration, restart Cursor completely. Then use in Agent mode:
Use tavily_search to find the latest AI developments
Use tavily_extract to get the content from https://example.com
Use tavily_search_qna to answer: What is the capital of France?
Available Tools
tavily_search
Search the web using Tavily API.
Parameters:
query(required): Search querysearch_depth: "basic" or "advanced" (default: "basic")topic: "general" or "news" (default: "general")days: Number of days back for news search (default: 3)max_results: Max results to return (default: 5, max: 20)include_images: Include images (default: false)include_answer: Include AI-generated answer (default: false)include_raw_content: Include raw HTML (default: false)
tavily_extract
Extract clean content from URLs.
Parameters:
urls(required): Array of URLs to extract from
tavily_search_qna
Get direct answers to questions.
Parameters:
query(required): The question to answersearch_depth: "basic" or "advanced" (default: "basic")
tavily_search_context
Generate context for RAG applications.
Parameters:
query(required): Search querysearch_depth: "basic" or "advanced" (default: "basic")max_results: Max results (default: 5)
Troubleshooting
Tools not showing up in Cursor
- Make sure you've completely quit and restarted Cursor (not just closed the window)
- Verify the path in
mcp.jsonis correct and absolute - Check that Node.js is installed:
node --version(should be >= 18.0.0) - Verify your Tavily API key is correct
"TAVILY_API_KEY environment variable is required" error
Make sure your API key is set in the env section of your mcp.json configuration.
Tools discovered but not usable
This was the original problem! This server fixes it by using underscores instead of hyphens in tool names.
Testing
You can test the server directly:
TAVILY_API_KEY=your-key-here node index.js
Then use the MCP Inspector or send MCP protocol messages via stdin.
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。