FetchSERP MCP Server

# Simple FetchSERP MCP Examples Get up and running with FetchSERP MCP in under 2 minutes! Choose your AI provider: ## Quick Start Options ### Option 1: Claude + MCP Server Run a local MCP server with Claude: ```bash # Create .env file first: echo "CLAUDE_API_KEY=your_claude_api_key" > .env echo "FETCHSERP_API_TOKEN=your_fetchserp_token" > .env echo "MCP_SERVER_URL=https://mcp.fetchserp.com/sse" > .env # Run the example: node simple-claude-mcp-example.js ``` ### Option 2: OpenAI + Hosted MCP Use OpenAI's hosted MCP integration: ```bash # Create .env file first: echo "OPENAI_API_KEY=your_openai_api_key" > .env echo "FETCHSERP_API_TOKEN=your_fetchserp_token" >> .env echo "MCP_SERVER_URL=https://mcp.fetchserp.com/sse" >> .env # Run the example: node simple-openai-mcp-example.js ``` **That's it!** Your AI will now use FetchSERP tools to get real search data. ## What These Examples Do ### Claude Example (`simple-claude-mcp-example.js`) 1. **Local MCP Server**: Runs FetchSERP MCP server locally 2. **Claude Integration**: Demonstrates Claude using real search tools via MCP ### OpenAI Example (`simple-openai-mcp-example.js`) 1. **Hosted MCP**: Uses OpenAI's hosted MCP integration with FetchSERP 2. **Approval Flow**: Handles the approval workflow for tool execution 3. **Polling Loop**: Waits for tool completion and displays results ## Example Output ### Claude Example ``` 🚀 Simple Claude API + MCP Server Example (HTTPS) ======================================================= 🤖 Asking Claude: "Who is ranking 3rd on Google for the keyword 'serp api'?" 📋 Claude's Response: 💬 I'll search for the keyword 'serp api' on Google to find out who is ranking 3rd. 🔧 Used MCP Tool: get_serp_results 📊 Tool Result: Success 💬 Based on the current Google search results for "serp api", **Bright Data** is ranking 3rd. ``` ### OpenAI Example ``` Approving tool request from the model... SERP results received! { data: { query: 'serp api', search_engine: 'google', country: 'us', results_count: 10, results: [ { site_name: 'serpapi.com', url: 'https://serpapi.com/', title: 'SerpApi: Google Search API', description: 'SerpApi is a real-time API to access Google search results...', ranking: 2 }, // ... more results ] } } ``` ## Requirements ### For Claude Example - Node.js 18+ - Claude API key ([get one here](https://console.anthropic.com/)) - FetchSERP API token ([get 250 free credits](https://www.fetchserp.com)) ### For OpenAI Example - Node.js 18+ - OpenAI API key ([get one here](https://platform.openai.com/)) - FetchSERP API token ([get 250 free credits](https://www.fetchserp.com)) ## Setup Instructions ### Claude Example Setup No additional setup needed - just run the command above. ### OpenAI Example Setup 1. Create a `.env` file: ```bash OPENAI_API_KEY=sk-your-openai-api-key-here FETCHSERP_API_TOKEN=fs_your-fetchserp-token-here MCP_SERVER_URL=https://mcp.fetchserp.com/sse ``` 2. Run the example: ```bash node simple-openai-mcp-example.js ``` ## Understanding the OpenAI MCP Flow The OpenAI example demonstrates the complete MCP approval workflow: 1. **Initial Request**: Model requests to use a tool 2. **Approval Required**: FetchSERP MCP returns an `mcp_approval_request` 3. **Auto-Approval**: Script automatically approves the request 4. **Tool Execution**: Tool runs and returns results 5. **Result Display**: Script parses and displays the search results This pattern can be adapted for any MCP tool that requires approval. ## Troubleshooting ### Claude Example Issues **Server not starting?** - Make sure you have Node.js 18+ installed - Check that port 8000 is available ### OpenAI Example Issues **Approval request errors?** - Make sure the `tools` array is included in both the initial and approval requests - Verify the `mcp_approval_response` format matches the expected structure ### General Issues **API errors?** - Verify your API keys are correct - Check you have sufficient API credits **No search results?** - Verify your FetchSERP API token is correct - Check you have API credits remaining at fetchserp.com ## What's Next? - Explore all 21+ available FetchSERP tools - Deploy to production with hosted MCP - Integrate with your own applications - Build custom MCP tools for your needs