Grants Search MCP Server

# Phase 2 Testing Guide - Grants MCP Server ## Setup Instructions ### 1. Configure API Key Replace `your_api_key_here` in the Claude Desktop config with your actual Simpler Grants API key: **File:** `/Users/tarive/Library/Application Support/Claude/claude_desktop_config.json` ```json "grantsmanship": { "command": "python3", "args": ["/Users/tarive/Desktop/grants-data-analysis/grants-mcp/main.py"], "env": { "SIMPLER_GRANTS_API_KEY": "your_actual_api_key_here" } } ``` ### 2. Restart Claude Desktop After updating the configuration: 1. Quit Claude Desktop completely (Cmd+Q) 2. Start Claude Desktop again 3. The MCP tools should appear in the tools list ## Available Tools After Phase 2 ### 1. **opportunity_discovery** (Original - Enhanced) Search and discover grant opportunities with advanced filtering. **Test Commands:** ``` Search for artificial intelligence grants Find renewable energy grants with award ceiling above $1 million Search for NSF grants in technology category ``` ### 2. **agency_landscape** (New in Phase 2) Map agencies and their funding focus areas with comprehensive analysis. **Test Commands:** ``` Analyze the agency landscape for top 5 agencies Show agency landscape focusing on NSF and NIH Map agency landscape for climate-related funding ``` ### 3. **funding_trend_scanner** (New in Phase 2) Scan funding trends to identify patterns and emerging opportunities. **Test Commands:** ``` Scan funding trends for the last 90 days Analyze funding trends in artificial intelligence category Show emerging topics and high-value opportunities in recent grants ``` ## Testing Scenarios ### Scenario 1: Basic Discovery ``` User: "Search for quantum computing grants" Expected: Tool uses opportunity_discovery to find relevant grants ``` ### Scenario 2: Agency Analysis ``` User: "Show me which agencies fund AI research" Expected: Tool uses agency_landscape to analyze agency portfolios ``` ### Scenario 3: Trend Analysis ``` User: "What are the emerging funding trends in the last month?" Expected: Tool uses funding_trend_scanner to identify patterns ``` ### Scenario 4: Combined Analysis ``` User: "I want to understand the renewable energy grants landscape" Expected: Multiple tools work together: 1. opportunity_discovery finds current grants 2. agency_landscape maps key agencies 3. funding_trend_scanner identifies trends ``` ## Verification Checklist - [ ] Claude Desktop restarted after config update - [ ] "grantsmanship" appears in available MCP tools - [ ] API key is correctly set (no 401 errors) - [ ] opportunity_discovery tool responds to queries - [ ] agency_landscape tool analyzes agencies - [ ] funding_trend_scanner identifies trends - [ ] Cache is working (second identical query is faster) ## Cache Behavior The optimized cache system uses different strategies: - **opportunity_discovery**: 5-minute cache (frequently changing) - **agency_landscape**: 1-hour cache (relatively stable) - **funding_trend_scanner**: 30-minute temporal buckets To test cache: 1. Run the same query twice 2. Second query should return instantly with "(from cache)" indicator ## Troubleshooting ### Tools Not Appearing 1. Check Claude Desktop logs: `~/Library/Logs/Claude/` 2. Verify Python path: `which python3` 3. Test server directly: `python3 /Users/tarive/Desktop/grants-data-analysis/grants-mcp/test_mcp_stdio.py` ### API Errors (401) 1. Verify API key is set correctly in config 2. Test API key: `export SIMPLER_GRANTS_API_KEY=your_key && python3 test_discovery_tools_api.py` ### Import Errors 1. Ensure you're in the project directory 2. Check Python version: `python3 --version` (should be 3.11+) 3. Verify imports: `python3 -c "from src.mcp_server.server import GrantsAnalysisServer"` ## Success Indicators ✅ All three tools respond to queries ✅ No errors in Claude Desktop logs ✅ Cache hit rate increases with repeated queries ✅ Results include detailed analysis and recommendations ✅ Cross-agency patterns are identified ✅ Emerging trends are detected ## Example Full Test Session ``` You: "Help me find funding opportunities for climate change research" Claude: [Uses opportunity_discovery to search] "I found 15 climate change research grants. Let me analyze the landscape..." [Uses agency_landscape to map agencies] "The main agencies funding climate research are EPA, NSF, and DOE..." [Uses funding_trend_scanner for trends] "I've identified increasing funding velocity in renewable energy with 45% growth..." You: "Show me the top high-value opportunities" Claude: [Returns cached trend data instantly] "Based on the analysis, here are the top 3 high-value opportunities..." ``` ## Phase 2 Features Summary | Feature | Tool | New/Enhanced | |---------|------|--------------| | Grant Search | opportunity_discovery | Enhanced | | Agency Portfolio Analysis | agency_landscape | New | | Cross-Agency Patterns | agency_landscape | New | | Temporal Trends | funding_trend_scanner | New | | Emerging Topics | funding_trend_scanner | New | | High-Value Opportunity Detection | funding_trend_scanner | New | | Optimized Caching | All tools | New | ## Next Steps After successful testing: 1. Document any issues found 2. Note performance improvements from caching 3. Identify most useful new features 4. Prepare for Phase 3 (Advanced Analytics) if desired