feat(mcp-server): Add create_project tool and improve test coverage

[jeremyeder]

Summary

Enhances the MCP Ambient Server with project creation capabilities and significantly improves test coverage.

Changes

API Client (client.py)

• ✅ Added create_project() method with proper error handling (ConnectError, TimeoutException)
• ✅ Fixed check_health() endpoint to correctly use /health (not under /api)

MCP Server (server.py)

• ✅ Added create_project MCP tool for creating projects via MCP interface
  • Parameters: name (required), display_name (optional), description (optional)
  • Returns JSON object with created project details

Test Suite (test_client.py)

• ✅ Fixed async mocking with pytest_asyncio and proper AsyncMock usage
• ✅ Added mock_httpx_response() helper for cleaner test setup
• ✅ Added 13 new test cases:
  • test_create_project_success / test_create_project_conflict
  • test_check_project_access_success / test_check_project_access_forbidden
  • test_get_session_success / test_get_session_not_found
  • test_get_session_k8s_resources_success
  • test_list_ootb_workflows_success
  • test_get_workflow_metadata_success / test_get_workflow_metadata_not_found
  • test_get_cluster_info_success
  • test_list_sessions_empty
  • test_list_session_workspace_success

E2E Tests (vteam.cy.ts)

• ✅ Added 3 integration tests verifying MCP server can reach backend endpoints:
  • Backend endpoint accessibility (/api/cluster-info, /api/workflows/ootb)
  • Authenticated project endpoints
  • Project creation workflow (UI + API verification)

Configuration (.gitignore)

• ✅ Added .mcp.json - Contains sensitive auth tokens, user-specific
• ✅ Added LOCAL_SETUP.md - Developer-specific setup documentation

Quality Assurance

All checks passing:

• ✅ Black formatting: All files properly formatted
• ✅ Flake8 linting: Clean (no E501, E203, W503 violations)
• ✅ pytest: 25/25 tests passing (1 minor warning, non-blocking)
• ✅ TypeScript: Syntax valid, no errors

Test Coverage

• Before: 12 test cases
• After: 25 test cases
• Improvement: +108% test coverage

Breaking Changes

None - all changes are additive.

Dependencies

No new dependencies added.

🤖 Generated with Claude Code

[github-actions]

Claude Code Review

Summary

This PR adds a new MCP (Model Context Protocol) server for the Ambient Code Platform API with comprehensive test coverage. The implementation introduces 14 MCP tools for read-only access to projects, sessions, and workflows, with one write operation (create_project). Code quality is excellent with proper async patterns, error handling, and security measures.

Issues by Severity

Blocker Issues

None - No blocking issues found.

Critical Issues

1. Token Security - Verified OK

• Status: VERIFIED - File uses environment variable placeholders, which is correct
• .mcp.json properly added to .gitignore

Major Issues

1. Missing Type Annotations in server.py:27

• Issue: api_client: APIClient = None uses implicit None type
• Fix: api_client: Optional[APIClient] = None

2. Broad Exception Catching

• Issue: Multiple except Exception blocks without specific exception types
• Fix: Catch specific exceptions (ValueError, httpx.HTTPError, etc.)

3. Health Endpoint URL Manipulation (client.py:352)

• Issue: String replacement for health URL is fragile
• Fix: Use proper URL parsing with urllib.parse

4. Missing Cleanup in Global Client

• Issue: api_client is initialized but never explicitly closed
• Fix: Add cleanup handler in main()

Minor Issues

• Inconsistent error message format
• Test coverage gap for get_health() method
• Documentation missing error handling examples
• Redundant error wrapping code

Positive Highlights

✅ Excellent Security Practices - Token length logging (PERFECT adherence to standards)
✅ Comprehensive Test Coverage - 25 test cases with proper async patterns
✅ Clean Code Architecture - Clear separation of concerns
✅ MCP Integration - Proper use of FastMCP framework

Recommendations

Priority 1 (Before Merge)

1. Fix type annotation: api_client: Optional[APIClient] = None
2. Add cleanup handler for api_client.close() in main()

Priority 2 (Follow-up PR)
3. Refactor health endpoint URL construction
4. Extract duplicate error handling
5. Add test_get_health_success
6. Tighten exception catching

Overall Assessment

Grade: A- (Excellent work with minor improvements needed)

Recommendation: APPROVE with Priority 1 fixes

---

🔍 View AI decision process (logs available for 90 days)

📋 View memory system files loaded (click to expand)

What Amber Loaded for Code Review

Amber automatically loaded these repository standards from the memory system:

1. CLAUDE.md - Master project instructions, development standards
2. backend-development.md - Go backend, K8s integration patterns
3. frontend-development.md - NextJS, Shadcn UI, React Query patterns
4. security-standards.md - Auth, RBAC, token handling
5. k8s-client-usage.md - User token vs service account patterns
6. error-handling.md - Consistent error patterns
7. react-query-usage.md - Data fetching patterns

Impact: This review used your repository's specific code quality standards, security patterns, and best practices from the memory system (PRs #359, #360) - not just generic code review guidelines.

[cgwalters]

Why just expose the abilty to create projects but nothing else?

The whole git commit message says "what" and not why which is far more important.

Is the high level goal here for an agentic session to be able to "recurse" on its own and spawn a full sub-workspace?