Reddit MCP Server

# MCP API Endpoint Audit & Missing Pieces ## Current MCP API Endpoints ### ✅ Courses - `POST /api/mcp/courses/batch` - Create multiple courses - `GET /api/mcp/courses` - List courses (public) - `GET /api/mcp/courses/my-courses` - Get user's courses - `GET /api/mcp/courses/[courseId]` - Get course details - `POST /api/mcp/courses/[courseId]/enroll` - Enroll in course - `GET /api/mcp/courses/[courseId]/progress` - Get course progress ### ✅ Units - `GET /api/mcp/courses/[courseId]/units` - Get course units - `POST /api/mcp/courses/[courseId]/units/batch` - Create multiple units ### ✅ Lessons - `GET /api/mcp/units/[unitId]/lessons` - Get unit lessons - `POST /api/mcp/units/[unitId]/lessons/batch` - Create multiple lessons - `GET /api/mcp/lessons/[lessonId]` - Get lesson details - `POST /api/mcp/lessons/[lessonId]/complete` - Mark lesson complete - `GET /api/mcp/lessons/search` - Search lessons ### ⚠️ Interactions (Partially) - `POST /api/mcp/lessons/[lessonId]/interaction` - Create interaction *(JUST ADDED)* - ❌ `GET /api/mcp/lessons/[lessonId]/interaction` - Get lesson interaction (MISSING) ### ❌ Prompts (ALL MISSING from MCP) - ❌ `POST /api/mcp/prompts` - Create single prompt - ❌ `POST /api/mcp/prompts/batch` - Create multiple prompts - ❌ `GET /api/mcp/lessons/[lessonId]/prompts` - Get lesson prompts - ❌ `GET /api/mcp/prompts/[promptId]` - Get prompt details - ❌ `PUT /api/mcp/prompts/[promptId]` - Update prompt - ❌ `DELETE /api/mcp/prompts/[promptId]` - Delete prompt ### ✅ Progress - `GET /api/mcp/progress` - Get user progress ### ✅ User - `GET /api/mcp/user/profile` - Get user profile - `GET /api/mcp/user/analytics` - Get user analytics --- ## Missing MCP Endpoints (Priority Order) ### 🔴 Critical - Needed for Questions/Interactions #### 1. **Prompts Management** ``` POST /api/mcp/prompts Create single prompt (with all fields) POST /api/mcp/prompts/batch Create multiple prompts at once GET /api/mcp/lessons/[lessonId]/prompts Get all prompts for a lesson (verification!) ``` **Why Critical:** - Claude needs to create questions/exercises after creating lessons - GET endpoint allows verification before creating duplicates - Reduces token waste by checking what exists #### 2. **Interaction Details** ``` GET /api/mcp/lessons/[lessonId]/interaction Get interaction details + ID ``` **Why Needed:** - Must get interaction ID before creating prompts - Current: Only POST exists to create, but no GET to retrieve existing - Prevents duplicate interaction creation ### 🟡 Important - For Better Reliability #### 3. **Verification/Check Endpoints** ``` GET /api/mcp/courses/[courseId]/check Check if course has lessons/units GET /api/mcp/units/[unitId]/check Check if unit has lessons GET /api/mcp/lessons/[lessonId]/check Check lesson completeness ``` **Why Important:** - Allows Claude to verify each step before proceeding - Reduces duplicate creation - Better error recovery #### 4. **Update Endpoints** (for fixing mistakes) ``` PUT /api/mcp/lessons/[lessonId] Update lesson content PUT /api/mcp/prompts/[promptId] Update prompt DELETE /api/mcp/prompts/[promptId] Delete prompt (to fix duplicates) ``` **Why Important:** - Fixing errors without recreating everything - Removing duplicates when they occur ### 🟢 Nice to Have - For Optimization #### 5. **Bulk Verification** ``` POST /api/mcp/verify/lesson-complete Check if lesson has content + prompts POST /api/mcp/verify/unit-complete Check if unit has all lessons ``` **Why Nice:** - Single API call to verify completeness - Reduces multiple round-trips - Token optimization #### 6. **Duplicate Detection** ``` GET /api/mcp/lessons/[lessonId]/duplicates Check for duplicate prompts POST /api/mcp/check-duplicates Batch check before creating ``` **Why Nice:** - Proactive duplicate prevention - Better UX for Claude --- ## Recommended Implementation Order ### Phase 1: Critical Prompts Support (DO THIS NOW) 1. ✅ `POST /api/mcp/lessons/[lessonId]/interaction` - Already created 2. ❌ `GET /api/mcp/lessons/[lessonId]/interaction` - Get interaction (critical!) 3. ❌ `GET /api/mcp/lessons/[lessonId]/prompts` - List prompts (verification!) 4. ❌ `POST /api/mcp/prompts` - Create single prompt 5. ❌ `POST /api/mcp/prompts/batch` - Create multiple prompts ### Phase 2: Update/Delete Support 6. ❌ `PUT /api/mcp/prompts/[promptId]` - Update prompt 7. ❌ `DELETE /api/mcp/prompts/[promptId]` - Delete prompt 8. ❌ `PUT /api/mcp/lessons/[lessonId]` - Update lesson ### Phase 3: Verification Helpers 9. ❌ `GET /api/mcp/lessons/[lessonId]/check` - Completeness check 10. ❌ `POST /api/mcp/verify/lesson-complete` - Bulk verification --- ## Current Workflow Issues ### Problem: Creating 5 Interactions per Lesson **Current Process:** 1. Create lesson → Get lesson ID 2. Create interaction → Get interaction ID 3. Create prompt 1 → Wait for response 4. Create prompt 2 → Wait for response 5. Create prompt 3 → Wait for response 6. Create prompt 4 → Wait for response 7. Create prompt 5 → Wait for response **Total: 7 API calls + 7 round trips per lesson** ### Proposed Optimized Process: 1. Create lesson → Get lesson ID 2. **Check if lesson has interaction** → Get interaction ID (NEW!) 3. **Check existing prompts** → Get count (NEW!) 4. Create missing prompts via batch → 1 call (NEW!) **Total: 4 API calls, prevents duplicates, much faster** --- ## Token Optimization Strategy ### Without Verification Endpoints: ``` Create lesson → Hope it worked → Create interaction → Hope it worked → Create 5 prompts individually → Hope they all worked → No way to check without creating duplicates ``` ### With Verification Endpoints: ``` Create lesson → Verify created → Get interaction (or create) → Check existing prompts → Create only missing prompts → Verify all 5 exist → Done! ``` **Tokens Saved:** - Prevents duplicate creation attempts - Reduces error recovery overhead - Single verification vs multiple create attempts --- ## Next Steps 1. **Immediate:** Create Phase 1 endpoints (prompts CRUD + interaction GET) 2. **Document:** Add MCP-specific docs for each endpoint 3. **Test:** Verify end-to-end flow works without duplicates 4. **Optimize:** Add batch verification endpoints --- ## API Design Principles for MCP ### ✅ DO: - Return full object details (IDs, counts, status) - Include verification data in responses - Make GET endpoints idempotent - Return existing resources instead of errors when appropriate - Include helpful metadata (count, exists, etc.) ### ❌ DON'T: - Require multiple calls to check existence - Return 404 when resource might exist elsewhere - Create duplicates silently - Make Claude guess if something worked --- ## Example Improved Response Format ### Bad (Current): ```json { "message": "Interaction created successfully", "interaction": { "id": "abc123" } } ``` ### Good (Proposed): ```json { "message": "Interaction created successfully", "interaction": { "id": "abc123", "lessonId": "lesson-1", "type": "assessment", "promptCount": 0, "createdAt": "2025-10-03T10:00:00Z" }, "lesson": { "id": "lesson-1", "title": "Introduction to Python", "hasInteraction": true, "promptCount": 0 } } ``` **Benefits:** - Claude knows interaction ID immediately - Can see if prompts already exist - Can verify lesson state in one call