feat: implement MCP server for AI assistant integration (#4132)

Implements #4129

## Summary
Adds MCP (Model Context Protocol) server integration to enable AI
assistants like Claude Code to discover plot specifications and retrieve
implementation code from the pyplots platform.

## Features
- **6 MCP tools** for plot discovery and code retrieval:
  - `list_specs` - Browse all specifications with pagination
- `search_specs_by_tags` - Filter by 9 tag categories (spec + impl
level)
  - `get_spec_detail` - Full spec with all library implementations
  - `get_implementation` - Single library code with quality score
  - `list_libraries` - All 9 supported plotting libraries
  - `get_tag_values` - Available values for tag discovery

## Technical Implementation
- FastMCP SDK integration
- Mounted at `/mcp` endpoint in FastAPI
- Reuses existing repositories (SpecRepository, LibraryRepository,
ImplRepository)
- Async database context management
- Response models from `api/schemas.py`

## Testing
- ✅ **15/15 unit tests passing** (with mocked database sessions)
- ✅ **Tested with live database** (253 specs, 9 libraries, all tools
working)
- ✅ **Code quality checks passing** (ruff lint + format)

## Dependencies Added
- `mcp>=1.0.0` (protocol package)
- `fastmcp>=0.1.0` (SDK)

## Commits
- `eac996e9` - Initial MCP server implementation
- `d05edb14` - Fix database model attribute names

Closes #4129

---------

Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: MarkusNeusinger

api/main.py

@@ -35,6 +35,7 @@
     stats_router,
 )
 from core.database import close_db, init_db, is_db_configured  # noqa: E402
+from pyplots_mcp.server import mcp_server  # noqa: E402
 
 
 # Configure logging
@@ -92,6 +93,7 @@ async def lifespan(app: FastAPI):
     allow_credentials=True,
     allow_methods=["*"],
     allow_headers=["*"],
+    expose_headers=["Mcp-Session-Id"],  # MCP session tracking
 )
 
 
@@ -123,6 +125,9 @@ async def add_cache_headers(request: Request, call_next):
     return response
 
 
+# Mount MCP server for AI assistant integration
+app.mount("/mcp", mcp_server.http_app())
+
 # Register routers
 app.include_router(health_router)
 app.include_router(stats_router)

pyplots_mcp/__init__.py

@@ -0,0 +1,5 @@
+"""
+MCP (Model Context Protocol) server for pyplots.
+
+Enables AI assistants to search plot specifications and fetch implementation code.
+"""