63.11 HANDOFF Cursor Mcp E2e Test

Documentation for Sessions

Handoff: MCP External Access End-to-End Test

Session: 63.11 Date: 2026-01-23 Handoff To: Cursor Agent (with browser capabilities) Priority: P0

Context

The MCP external access API has been deployed and verified working on Railway. The endpoints are responding correctly:

Endpoint	Status	Auth
`GET /ai-api/mcp/health`	Operational	Public
`GET /ai-api/mcp/openapi.yaml`	Working	Public
`POST /ai-api/mcp/preflight`	401	Required
`POST /ai-api/mcp/tools/call`	401	Required
`GET /ai-api/mcp/tools`	401	Required
`GET /ai-api/mcp/agents`	401	Required

What's missing: We need to test the complete flow with actual authentication.

Your Mission

Test the complete MCP external access flow end-to-end:

Step 1: Create MCP Project (via GoGrab Space UI)

Navigate to https://gograb.space
Log in (or create account if needed)
Go to AI Hub → MCP Projects (or similar navigation)
Create a new MCP project
Capture the API Key - it's only shown once!
Note the Project ID

Step 2: Test Authenticated Endpoints

Use the captured credentials to test:

# Set your credentials
API_KEY="mcp_your_api_key_here"
PROJECT_ID="your_project_id_here"

# Test 1: Get available tools
curl -X GET "https://ai-api.gograb.space/ai-api/mcp/tools" \
  -H "X-MCP-API-Key: $API_KEY" \
  -H "X-MCP-Project-Id: $PROJECT_ID"

# Test 2: Preflight (get context for a task)
curl -X POST "https://ai-api.gograb.space/ai-api/mcp/preflight" \
  -H "Content-Type: application/json" \
  -H "X-MCP-API-Key: $API_KEY" \
  -H "X-MCP-Project-Id: $PROJECT_ID" \
  -d '{"task": "Testing MCP external access"}'

# Test 3: Log a decision
curl -X POST "https://ai-api.gograb.space/ai-api/mcp/tools/call" \
  -H "Content-Type: application/json" \
  -H "X-MCP-API-Key: $API_KEY" \
  -H "X-MCP-Project-Id: $PROJECT_ID" \
  -d '{
    "tool": "log_decision",
    "arguments": {
      "type": "implementation",
      "question": "Testing MCP external access",
      "chosen": "Verified API endpoints working",
      "reasoning": "End-to-end test from Cursor agent",
      "confidence": 0.9
    }
  }'

# Test 4: Find similar decisions
curl -X POST "https://ai-api.gograb.space/ai-api/mcp/tools/call" \
  -H "Content-Type: application/json" \
  -H "X-MCP-API-Key: $API_KEY" \
  -H "X-MCP-Project-Id: $PROJECT_ID" \
  -d '{
    "tool": "find_similar",
    "arguments": {
      "query": "MCP external access"
    }
  }'

# Test 5: Get agents list
curl -X GET "https://ai-api.gograb.space/ai-api/mcp/agents" \
  -H "X-MCP-API-Key: $API_KEY" \
  -H "X-MCP-Project-Id: $PROJECT_ID"

# Test 6: Get usage stats
curl -X GET "https://ai-api.gograb.space/ai-api/mcp/usage" \
  -H "X-MCP-API-Key: $API_KEY" \
  -H "X-MCP-Project-Id: $PROJECT_ID"

Step 3: Verify Project Database Isolation

After logging a decision, verify it was stored in the project's isolated database:

The decision should appear in find_similar results
Usage stats should increment

Step 4: Test OpenAI Functions Export

curl -X GET "https://ai-api.gograb.space/ai-api/mcp/openai/functions" \
  -H "X-MCP-API-Key: $API_KEY" \
  -H "X-MCP-Project-Id: $PROJECT_ID"

This should return OpenAI-compatible function definitions.

Expected Results

Test	Expected Result
Create Project	Project created, API key generated
GET /tools	List of available MCP tools
POST /preflight	Context with patterns, conventions
POST /tools/call (log_decision)	Decision logged successfully
POST /tools/call (find_similar)	Returns logged decision
GET /agents	List of agent experts
GET /usage	Usage statistics for project
GET /openai/functions	OpenAI function definitions

If Something Fails

401 Unauthorized

Verify API key format: should start with mcp_
Verify Project ID matches the project
Check if API key was regenerated

404 Not Found

Endpoint might not exist
Check the OpenAPI spec: https://ai-api.gograb.space/ai-api/mcp/openapi.yaml

500 Server Error

Check Railway logs for backend-ai-api
Report the error message in the test results

Report Template

After testing, create a report at: docs/english/sessions/editor2/session-63/63.12-TEST-RESULTS-mcp-e2e.md

Include:

Project creation: Success/Fail
Each endpoint test result
Any errors encountered
Screenshots if relevant
Recommendations for fixes

Related Files

Controller: apis/backend-ai-api/src/app/mcp-gateway/mcp-gateway.controller.ts
Auth Guard: apis/backend-ai-api/src/app/mcp-gateway/guards/jwt-or-api-key.guard.ts
Tools Bridge: apis/backend-ai-api/src/app/mcp-gateway/services/mcp-tools-bridge.service.ts
OpenAPI Spec: apis/backend-ai-api/src/app/mcp-gateway/openapi/mcp-openapi.yaml

Success Criteria

MCP project created via UI
API key works for authentication
All tool endpoints respond correctly
Decisions are stored in project database
Usage tracking increments
OpenAI functions export works

Handoff created by Claude Code. Ready for Cursor agent pickup.