Umbrella MCP Server
find-get-all-recommendations.cjs•7.97 kB
#!/usr/bin/env node
/**
* UMBRELLA MCP SERVER: get_all_recommendations TOOL ANALYSIS
*
* This script analyzes where and how the `get_all_recommendations` tool is handled in the MCP server code.
* Based on user feedback:
* - For direct customer accounts: get_all_recommendations works perfectly fine
* - For MSP accounts: The tool fails because it needs customer detection from natural language queries
*/
console.log('='.repeat(80));
console.log('UMBRELLA MCP SERVER: get_all_recommendations TOOL ANALYSIS');
console.log('='.repeat(80));
console.log('\n🔍 FINDINGS SUMMARY:');
console.log('-------------------');
console.log('\n1. TOOL REGISTRATION:');
console.log(' ✅ Tool is dynamically registered from UMBRELLA_ENDPOINTS array');
console.log(' 📍 Location: /Users/david/Downloads/MCP/UmbrellaMCP/src/types.ts:192-203');
console.log(' 📄 Endpoint Definition:');
console.log(' {');
console.log(' path: "/api/v2/recommendations/list",');
console.log(' method: "POST",');
console.log(' description: "Get all recommendations with pagination support. Returns detailed recommendations with savings potential",');
console.log(' category: "Recommendations",');
console.log(' parameters: {');
console.log(' "filters": "JSON object with user_status, status_filter, and date ranges",');
console.log(' "sort": "Array of sort objects with by and order fields",');
console.log(' "page_size": "Number of recommendations per page (default 2000)",');
console.log(' "pagination_token": "Token for fetching next page of results",');
console.log(' "userQuery": "Natural language query from user - used for customer detection"');
console.log(' }');
console.log(' }');
console.log('\n2. TOOL NAME GENERATION:');
console.log(' 📍 Location: /Users/david/Downloads/MCP/UmbrellaMCP/src/server.ts:107-129');
console.log(' 🔧 Algorithm: api__${path.replace(/\\//g, \'_\').replace(/[-]/g, \'_\')}');
console.log(' 📝 Expected Tool Name: "api__api_v2_recommendations_list"');
console.log(' ❌ NOT "get_all_recommendations" (user might be using wrong name)');
console.log('\n3. TOOL CALL HANDLING:');
console.log(' 📍 Location: /Users/david/Downloads/MCP/UmbrellaMCP/src/server.ts:196-218');
console.log(' 🔧 Flow: name.startsWith("api__") → find endpoint → handleApiCall()');
console.log(' 📍 handleApiCall: /Users/david/Downloads/MCP/UmbrellaMCP/src/server.ts:535-1235');
console.log('\n4. MSP CUSTOMER DETECTION:');
console.log(' 📍 Location: /Users/david/Downloads/MCP/UmbrellaMCP/src/server.ts:764-821');
console.log(' 🔧 Method: detectCustomerFromQuery()');
console.log(' 📍 Implementation: /Users/david/Downloads/MCP/UmbrellaMCP/src/server.ts:1270-1546');
console.log(' ✅ ALREADY IMPLEMENTED for recommendation endpoints');
console.log(' 🎯 Detection Logic:');
console.log(' - Only for MSP users (cognito auth method)');
console.log(' - Only for recommendation/cost-related endpoints (includes /recommendations)');
console.log(' - Analyzes userQuery parameter for customer names');
console.log(' - Maps customer names to account keys and division IDs');
console.log('\n5. API CLIENT FLOW:');
console.log(' 📍 Location: /Users/david/Downloads/MCP/UmbrellaMCP/src/api-client.ts:84-97');
console.log(' 🔧 General Flow: makeRequest() → makeDirectRequest()');
console.log(' 📍 Special Recommendations Method: /Users/david/Downloads/MCP/UmbrellaMCP/src/api-client.ts:412-558');
console.log(' 🔧 Dedicated getRecommendations() method with:');
console.log(' - Hardcoded V2 API endpoint: https://api.umbrellacost.io/api/v2/recommendations/list');
console.log(' - POST method with specific request body');
console.log(' - Pagination support');
console.log(' - 12-month date range default');
console.log('\n6. THE ISSUE ANALYSIS:');
console.log(' 🤔 Based on the code analysis, the MSP customer detection is ALREADY implemented.');
console.log(' 🧐 Possible causes of MSP account failures:');
console.log('\n A. TOOL NAME MISMATCH:');
console.log(' ❌ User calls: "get_all_recommendations"');
console.log(' ✅ Actual name: "api__api_v2_recommendations_list"');
console.log(' 💡 Solution: Use correct tool name or add alias');
console.log('\n B. MISSING userQuery PARAMETER:');
console.log(' 🎯 Customer detection requires "userQuery" parameter');
console.log(' 📝 Example: userQuery="Show me Bank Leumi recommendations"');
console.log(' 💡 Solution: Ensure Claude Desktop passes userQuery');
console.log('\n C. AUTHENTICATION METHOD CHECK:');
console.log(' 🔍 Detection only works for cognito auth (MSP users)');
console.log(' 📍 Check: session.auth.getUserManagementInfo()?.authMethod === "cognito"');
console.log(' 💡 Solution: Verify MSP account uses cognito auth');
console.log('\n D. API FLOW CONFLICT:');
console.log(' 🔄 Two paths to recommendations:');
console.log(' 1. Generic: makeRequest("/api/v2/recommendations/list") via handleApiCall()');
console.log(' 2. Dedicated: getRecommendations() method');
console.log(' ⚠️ The dedicated method bypasses MCP customer detection logic!');
console.log('\n7. ROOT CAUSE IDENTIFIED:');
console.log(' 🎯 THE PROBLEM: Dedicated getRecommendations() method bypasses MCP server logic');
console.log(' 📍 Location: /Users/david/Downloads/MCP/UmbrellaMCP/src/api-client.ts:412-558');
console.log(' 🔧 This method:');
console.log(' - Makes direct axios.post() call to hardcoded URL');
console.log(' - Bypasses makeDirectRequest() and MCP customer detection');
console.log(' - Only uses getAuthHeaders() without customer context');
console.log(' - Works for direct customers (single account)');
console.log(' - Fails for MSP customers (needs account switching)');
console.log('\n8. SOLUTION OPTIONS:');
console.log('\n OPTION 1: Ensure MCP Tool Uses makeRequest (RECOMMENDED)');
console.log(' 📍 Problem: The api__api_v2_recommendations_list tool should use makeRequest(), not getRecommendations()');
console.log(' 🔧 Check server.ts:927 - does it call session.apiClient.makeRequest() or getRecommendations()?');
console.log(' ✅ If it uses makeRequest(), customer detection will work');
console.log('\n OPTION 2: Add Customer Detection to getRecommendations()');
console.log(' 📍 Modify: /Users/david/Downloads/MCP/UmbrellaMCP/src/api-client.ts:412-558');
console.log(' 🔧 Add customer_account_key parameter handling');
console.log(' 🔧 Extract customer info from params before making axios call');
console.log(' 🔧 Include customer context in auth headers');
console.log('\n OPTION 3: Add Tool Name Alias');
console.log(' 📍 Modify: /Users/david/Downloads/MCP/UmbrellaMCP/src/server.ts');
console.log(' 🔧 Add special case for "get_all_recommendations" → "api__api_v2_recommendations_list"');
console.log('\n9. VERIFICATION STEPS:');
console.log(' 1. Check which API client method is called by the MCP tool');
console.log(' 2. Verify MSP account uses cognito authentication');
console.log(' 3. Ensure userQuery parameter is passed from Claude Desktop');
console.log(' 4. Test customer detection with debug logging');
console.log(' 5. Verify auth headers include customer context');
console.log('\n10. DEBUG COMMANDS:');
console.log(' # Check actual tool call in server logs');
console.log(' grep "api__api_v2_recommendations_list" server_logs.txt');
console.log('');
console.log(' # Check customer detection logs');
console.log(' grep "CUSTOMER-DETECTION" server_logs.txt');
console.log('');
console.log(' # Check auth method');
console.log(' grep "getUserManagementInfo" server_logs.txt');
console.log('\n' + '='.repeat(80));
console.log('NEXT STEPS: Examine server.ts:927 to see which API client method is called');
console.log('='.repeat(80));
Latest Blog Posts
- Federated Learning with MCP: Building Privacy-Preserving Agents Across Distributed EdgesBy Om-Shree-0709 on .SecuremcpLearning
- MCP Moves to the Linux Foundation: Neutral Stewardship for Agentic InfrastructureBy Om-Shree-0709 on .mcpanthropicLinux Foundation
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/daviddraiumbrella/invoice-monitoring'
If you have feedback or need assistance with the MCP directory API, please join our Discord server