Tavily MCP Server
# Tavily MCP Server
A production-ready MCP (Model Context Protocol) server that provides web search capabilities using the Tavily API. This server integrates seamlessly with Roo and other MCP-compatible AI assistants.
## Features
- ๐ **Web Search**: Powerful web search using Tavily's AI-optimized search API
- ๐ฏ **Direct Answers**: Get immediate answers to queries when available
- ๐ **Configurable Results**: Control search depth, result count, and domain filtering
- ๐ **Production Ready**: Built with TypeScript, comprehensive testing, and PM2 deployment
- ๐ **Secure**: Environment-based API key management
- ๐ **Monitoring**: Full logging and process monitoring with PM2
- ๐งช **Well Tested**: Comprehensive unit and integration test coverage
## Quick Start
### Prerequisites
- Node.js 18+
- npm or yarn
- Tavily API key ([Get one here](https://tavily.com))
- PM2 (for production deployment)
### Installation & Deployment
1. **Clone and setup:**
```bash
cd tavily-mcp-server
npm install
```
2. **Set your API key:**
```bash
export TAVILY_API_KEY="your-api-key-here"
```
3. **Run tests:**
```bash
npm test
npm run test:coverage
```
4. **Deploy with PM2:**
```bash
./deploy.sh
```
That's it! The server is now running and ready for MCP connections.
## Development
### Build and Test
```bash
# Install dependencies
npm install
# Run in development mode
npm run dev
# Build for production
npm run build
# Run unit tests
npm test
# Run tests with coverage
npm run test:coverage
# Run integration tests
./test-mcp.js
# Lint code
npm run lint
npm run lint:fix
```
### Testing
The project includes comprehensive testing:
- **Unit Tests**: Test individual components and functions
- **Integration Tests**: Test the complete MCP server functionality
- **MCP Protocol Tests**: Validate MCP protocol compliance
- **API Tests**: Test Tavily API integration (requires valid API key)
```bash
# Run all tests
npm test
# Run with coverage report
npm run test:coverage
# Test the actual MCP server
./test-mcp.js
```
## Configuration
### Environment Variables
- `TAVILY_API_KEY` (required): Your Tavily API key
- `NODE_ENV` (optional): Set to "production" for production deployment
### PM2 Configuration
The `pm2-apps.json` file contains production configuration:
```json
{
"apps": [{
"name": "tavily-mcp-server",
"script": "dist/index.js",
"instances": 1,
"exec_mode": "fork",
"env": {
"NODE_ENV": "production",
"TAVILY_API_KEY": "your-api-key"
}
}]
}
```
## Usage with Roo
### Global Installation
Add to your global MCP settings (`~/.roo/mcp_settings.json`):
```json
{
"mcpServers": {
"tavily-search": {
"command": "node",
"args": ["/home/ubuntu/roo-tavily/tavily-mcp-server/dist/index.js"],
"env": {
"TAVILY_API_KEY": "your-api-key-here"
}
}
}
}
```
### Project-specific Installation
Add to your project's MCP settings (`.roo/mcp.json`):
```json
{
"mcpServers": {
"tavily-search": {
"command": "node",
"args": ["./tavily-mcp-server/dist/index.js"],
"env": {
"TAVILY_API_KEY": "your-api-key-here"
}
}
}
}
```
### Using the Web Search Tool
Once configured, you can use the web search tool in Roo:
```xml
<use_mcp_tool>
<server_name>tavily-search</server_name>
<tool_name>web_search</tool_name>
<arguments>
{
"query": "latest developments in AI",
"search_depth": "advanced",
"max_results": 10,
"include_answer": true
}
</arguments>
</use_mcp_tool>
```
## API Reference
### web_search Tool
Search the web using Tavily's AI-optimized search API.
#### Parameters
| Parameter | Type | Required | Default | Description |
|-----------|------|----------|---------|-------------|
| `query` | string | โ
| - | The search query to execute |
| `search_depth` | string | โ | "basic" | Search depth: "basic" or "advanced" |
| `include_answer` | boolean | โ | true | Whether to include a direct answer |
| `max_results` | number | โ | 5 | Number of results (1-20) |
| `include_domains` | string[] | โ | - | Domains to include in search |
| `exclude_domains` | string[] | โ | - | Domains to exclude from search |
#### Response Format
The tool returns formatted search results including:
- **Direct Answer**: AI-generated answer to the query (if available)
- **Search Results**: List of relevant web pages with:
- Title and URL
- Content snippet
- Relevance score
- Publication date (if available)
- **Follow-up Questions**: Suggested related queries
#### Example Response
```
# Search Results for: "latest developments in AI"
## Direct Answer
Recent AI developments include advances in large language models,
multimodal AI systems, and improved reasoning capabilities...
## Search Results
### 1. Major AI Breakthroughs in 2024
**URL:** https://example.com/ai-breakthroughs
**Published:** 2024-01-15
**Score:** 0.95
Recent developments in artificial intelligence have shown remarkable
progress in areas such as natural language processing...
---
### 2. OpenAI Announces GPT-5
**URL:** https://example.com/gpt5-announcement
**Score:** 0.92
OpenAI has announced the development of GPT-5, promising significant
improvements in reasoning and multimodal capabilities...
---
## Follow-up Questions
1. What are the implications of these AI developments?
2. How do these advances compare to previous years?
3. What challenges remain in AI development?
```
## Production Deployment
### PM2 Management
```bash
# Start the server
pm2 start pm2-apps.json
# View status
pm2 status
# View logs
pm2 logs tavily-mcp-server
# Restart server
pm2 restart tavily-mcp-server
# Stop server
pm2 stop tavily-mcp-server
# Monitor all processes
pm2 monit
```
### Nginx Reverse Proxy (Optional)
If you need HTTP access, you can set up an Nginx reverse proxy:
```nginx
server {
listen 80;
server_name your-domain.com;
location / {
proxy_pass http://localhost:3000;
proxy_http_version 1.1;
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Connection 'upgrade';
proxy_set_header Host $host;
proxy_cache_bypass $http_upgrade;
}
}
```
### Monitoring and Logs
- **Application Logs**: `/var/log/pm2/tavily-mcp-server.log`
- **Error Logs**: `/var/log/pm2/tavily-mcp-server-error.log`
- **PM2 Monitoring**: `pm2 monit`
## Troubleshooting
### Common Issues
1. **"TAVILY_API_KEY environment variable is required"**
- Ensure your API key is set: `export TAVILY_API_KEY="your-key"`
- Check PM2 config has the correct API key
2. **"Cannot find module" errors**
- Run `npm install` to install dependencies
- Ensure you've built the project: `npm run build`
3. **Server won't start**
- Check logs: `pm2 logs tavily-mcp-server`
- Verify API key is valid
- Ensure port is not in use
4. **Search requests failing**
- Verify API key is valid and has credits
- Check network connectivity
- Review error logs for specific API errors
### Debug Mode
Run the server in debug mode:
```bash
NODE_ENV=development npm run dev
```
### Testing Connection
Test the MCP server directly:
```bash
./test-mcp.js
```
## Contributing
1. Fork the repository
2. Create a feature branch: `git checkout -b feature-name`
3. Make your changes
4. Add tests for new functionality
5. Ensure all tests pass: `npm test`
6. Submit a pull request
## License
MIT License - see LICENSE file for details.
## Support
- ๐ง Email: support@roo.com
- ๐ Issues: [GitHub Issues](https://github.com/roo/tavily-mcp-server/issues)
- ๐ Documentation: [Roo Documentation](https://docs.roo.com)
---
Built with โค๏ธ by the Roo teamTDQS
Scored across 1 tool
With only one tool, there is no possibility of confusion or overlap between tools. The single tool 'web_search' has a clearly defined and distinct purpose, making it impossible for an agent to misselect between non-existent alternatives.
Since there is only one tool, naming consistency is inherently perfect. The tool name 'web_search' follows a clear verb_noun pattern, but with no other tools to compare against, there can be no inconsistency in the set.
A single tool is too few for a server named 'Tavily MCP Server', which implies broader web search capabilities. While 'web_search' covers the core function, typical search servers might include tools for advanced queries, filtering, or result analysis, making this feel thin and limited in scope.
The tool surface is severely incomplete for a web search domain. It lacks obvious gaps such as tools for refining searches (e.g., by date or source), handling pagination, or accessing cached content, which could lead to agent failures when trying to perform comprehensive search tasks.