SkyFi MCP Server

# SkyFi MCP Integration Guide How to connect AI agents to the SkyFi MCP server. ## Overview The SkyFi MCP server implements the Model Context Protocol, allowing AI agents like Claude and GPT to access satellite imagery services through a standardized interface. ## Claude Desktop Integration ### Configuration Add to your Claude Desktop config file: **macOS:** `~/Library/Application Support/Claude/claude_desktop_config.json` **Windows:** `%APPDATA%\Claude\claude_desktop_config.json` ```json { "mcpServers": { "skyfi": { "command": "node", "args": ["/path/to/SkyFiMCP/dist/index.js"], "env": { "SKYFI_API_KEY": "your-api-key" } } } } ``` ### Using with npx ```json { "mcpServers": { "skyfi": { "command": "npx", "args": ["-y", "skyfi-mcp"], "env": { "SKYFI_API_KEY": "your-api-key" } } } } ``` ### Verify Connection After restarting Claude Desktop, you should see the SkyFi tools available. Try asking: > "Search for satellite imagery of San Francisco from the last month" ## Claude Code Integration ### Configuration Create or edit `.mcp.json` in your project root: ```json { "mcpServers": { "skyfi": { "command": "node", "args": ["/path/to/SkyFiMCP/dist/index.js"], "env": { "SKYFI_API_KEY": "your-api-key" } } } } ``` ### Usage The tools will be available as `mcp__skyfi__<tool_name>`. For example: ``` mcp__skyfi__search_archive mcp__skyfi__get_pricing_estimate mcp__skyfi__place_order ``` ## OpenAI GPT Integration ### Custom GPT Actions You can expose the MCP server as an API for custom GPTs: 1. Run the MCP server with HTTP transport 2. Create an OpenAPI spec for the tools 3. Add as an action in your custom GPT ### Function Calling For direct API integration: ```python import openai tools = [ { "type": "function", "function": { "name": "search_archive", "description": "Search for satellite imagery", "parameters": { "type": "object", "properties": { "location": { "type": "string", "description": "Location to search" }, "resolution": { "type": "string", "description": "Minimum resolution" } }, "required": ["location"] } } } ] response = openai.chat.completions.create( model="gpt-4", messages=[{"role": "user", "content": "Find satellite imagery of NYC"}], tools=tools ) ``` ## Common Workflows ### 1. Search and Order Workflow ``` User: Find high-resolution imagery of downtown Seattle Agent: 1. search_archive(location="downtown Seattle", resolution="0.5m") 2. Present results to user 3. get_pricing_estimate(imageId="selected_image_id") 4. User confirms price 5. place_order(imageId="selected_image_id", userConfirmationToken="token") 6. poll_order_status(orderId="order_id") 7. Return download URLs when complete ``` ### 2. New Capture Workflow ``` User: I need new imagery of this construction site Agent: 1. check_order_feasibility(taskingRequest={location, resolution, captureDate}) 2. If feasible, get_pricing_estimate(taskingRequest={...}) 3. User confirms price 4. place_order(taskingRequest={...}, userConfirmationToken="token") 5. poll_order_status to track progress ``` ### 3. Monitoring Workflow ``` User: Alert me when new imagery is available for my farm Agent: 1. create_monitor(location={polygon}, webhookUrl="...", frequency="daily") 2. Return monitor ID and confirmation 3. User receives webhook notifications when new imagery matches ``` ## Best Practices ### API Key Security - Never hardcode API keys in source code - Use environment variables - Rotate keys periodically - Use separate keys for development and production ### Error Handling Always handle potential errors: ```javascript const result = await searchArchive({ apiKey: process.env.SKYFI_API_KEY, location: "San Francisco" }); if (result.isError) { // Handle error console.error(result.content[0].text); } else { // Process results const data = JSON.parse(result.content[0].text); } ``` ### Rate Limiting - Implement exponential backoff for retries - Cache results when appropriate - Use polling with reasonable intervals (30+ seconds) ### User Confirmation For order placement, always: 1. Show pricing to user first 2. Get explicit confirmation 3. Use the confirmation token 4. Never auto-approve orders without user consent ## Webhook Integration ### Webhook Payload When a monitor triggers, SkyFi sends: ```json { "event": "new_imagery", "monitorId": "mon_abc123", "imagery": { "id": "img_xyz789", "captureDate": "2025-01-15T10:00:00Z", "resolution": "1m", "price": 45.00 }, "timestamp": "2025-01-15T10:30:00Z" } ``` ### Webhook Security - Verify webhook signatures - Use HTTPS endpoints - Implement idempotency for duplicate events ## Testing ### Mock Server For development and testing, you can use the test fixtures: ```javascript import nock from 'nock'; nock('https://api.skyfi.com/v1') .post('/archive/search') .reply(200, { results: [...], total: 10, hasMore: false }); ``` ### Test API Key Contact SkyFi support for a sandbox API key with test credentials. ## Troubleshooting See [Deployment Guide - Troubleshooting](../deployment/README.md#troubleshooting) for common issues and solutions.