Skip to main content

Testing

Testing strategy and manual testing procedures for MCP Gatekeeper.

Current Testing Status​

MCP Gatekeeper currently uses manual testing. Automated test suites are planned for future development.

Manual Testing​

Backend Testing​

API Endpoints​

Test using the interactive API docs:

  1. Start backend: npm run dev
  2. Open http://127.0.0.1:8000/docs
  3. Test each endpoint with sample data
  4. Verify response codes and data

HTTP Proxy Testing​

# Test proxy with curl
curl -X POST http://127.0.0.1:8000/proxy/test_agent/github \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"initialize","id":1,"params":{}}'

Database Testing​

Verify database operations:

cd apps/backend
sqlite3 mcpgatekeeper.db
# Check tables
.tables
# Query data
SELECT * FROM http_mcp_servers;

Frontend Testing​

UI Components​

  1. Navigation: Test all sidebar links
  2. Forms: Submit with valid/invalid data
  3. Tables: Sort, filter, pagination
  4. Modals: Open, close, form submission
  5. Theme Toggle: Switch between light/dark

Agent Configuration​

  1. Add server to each agent type
  2. Verify config file updated
  3. Test server removal
  4. Check for file permission errors

HTTP Server Management​

  1. Add HTTP server
  2. Test connection
  3. Edit server configuration
  4. Add to agent
  5. Verify proxy URL format
  6. Delete server

Integration Testing​

STDIO Server Flow​

  1. Install package from npm
  2. Add to agent configuration
  3. Trigger agent to use server
  4. Verify metrics appear in UI
  5. Check logs are recorded

HTTP Server Flow​

  1. Add HTTP server (e.g., GitHub MCP)
  2. Add to agent configuration
  3. Agent connects through proxy
  4. Verify proxy logs in backend
  5. Check metrics in UI
  6. Verify request/response logged

Cross-platform Testing​

Test on:

  • macOS: Primary development platform
  • Windows: Verify path handling
  • Linux: Verify permissions

Agent Compatibility Testing​

Test with each supported agent:

  • Windsurf: Uses serverUrl
  • Amazon Q: Uses url
  • Gemini CLI: Uses httpUrl
  • Claude Desktop: Uses url

Browser Testing​

Test frontend in:

  • Chrome/Edge (Chromium)
  • Firefox
  • Safari

Test Scenarios​

Critical Path​

  1. Start application
  2. Install STDIO package
  3. Add HTTP server
  4. Configure agent with both
  5. Verify metrics appear
  6. Check logs are visible
  7. Test theme switching

Error Scenarios​

  1. Invalid package name
  2. Network timeout for HTTP server
  3. Permission denied on config file
  4. Invalid JSON in requests
  5. Missing authentication headers

Performance Testing​

  1. Install multiple packages
  2. Configure multiple HTTP servers
  3. Generate high metric volume
  4. Verify UI responsiveness
  5. Check database query performance

Debugging​

Backend Logs​

# Backend console shows:
# - HTTP proxy requests/responses
# - Database operations
# - Error stack traces

Frontend Console​

// Browser DevTools Console shows:
// - API calls
// - WebSocket messages
// - React errors

Database Inspection​

# Check metrics
sqlite3 apps/backend/mcpgatekeeper.db "SELECT * FROM http_server_metrics ORDER BY timestamp DESC LIMIT 10;"

# Check logs
sqlite3 apps/backend/mcpgatekeeper.db "SELECT * FROM server_logs ORDER BY timestamp DESC LIMIT 10;"

Future Testing Plans​

  • Unit tests: Backend API endpoints, service logic
  • E2E tests: Full user workflows with Playwright/Cypress
  • Integration tests: Database operations, proxy functionality
  • Load tests: High-volume metrics collection

For contributing test cases, see Contributing.