┌─────────────────┐     ┌─────────────────┐
│   MCP Clients   │     │   Web Clients   │
│  (Claude Code)  │     │   (Browser)     │
└────────┬────────┘     └────────┬────────┘
         │                       │
         ▼                       ▼
┌─────────────────┐     ┌─────────────────┐
│   MCP Server    │     │  FastAPI Server │
│   (mcp_server)  │     │   (rest_api)    │
│  - SSE/HTTP     │     │  - RESTful API  │
└────────┬────────┘     └────────┬────────┘
         │                       │
         └───────────┬───────────┘
                     ▼
         ┌─────────────────────┐
         │   Business Server   │
         │   (business)        │
         │  - Core business    │
         │  - Data storage     │
         │  - Tag system       │
         └─────────────────────┘

ai_memory_mcp/
├── src/
│   ├── business/      # Business logic layer (core)
│   ├── mcp_server/    # MCP server layer
│   ├── rest_api/      # FastAPI REST API layer
│   ├── clients/       # Client modules
│   └── common/        # Common modules
├── docker/            # Docker deployment files
├── test/              # Test files (unit/integration/e2e/performance)
├── scripts/           # Utility scripts
├── examples/          # Example code (agents/skills)
├── docs/              # Documentation
├── config/            # Configuration files
├── scripts/           # Utility scripts
│   ├── start_mcp.py         # MCP startup script
│   ├── start_business.py    # Business service startup script
│   └── start_fastapi.py     # FastAPI startup script

# 1. Enter Docker directory
cd docker

# 2. Start service (default port 8000)
./manage.sh start

# 3. Check status
./manage.sh status

# 4. Stop service
./manage.sh stop

# 5. View logs
docker logs -f ai-memory-mcp

# 1. Create Conda environment
conda create -n ai_memory_mcp python=3.12
conda activate ai_memory_mcp

# 2. Install dependencies
pip install -r requirements.txt

# 3. Start business service
python scripts/start_business.py

# 4. Start MCP server (new terminal)
python scripts/start_mcp.py

# 5. Start FastAPI service (optional, new terminal)
python scripts/start_fastapi.py

{
  "mcpServers": {
    "memory": {
      "url": "http://localhost:8000/mcp",
      "type": "http"
    }
  }
}

Please use MCP [memory] to list all projects

┌─────────────────┐
│   Features      │ ─┐
│   (Feature List)│  │ Associated
└─────────────────┘  │
                     ├─→ Notes (Notes)
┌─────────────────┐  │    (Record development process/decision rationale)
│   Fixes         │ ─┘
│   (Bug Fixes)   │
└─────────────────┘

You: Continue improving API authentication, add refresh token mechanism

Claude: [Query memory] Found project records:
        - feat_20260401: JWT authentication implemented using FastAPI Security
        - note_20260401: Key stored at /secrets/jwt_key.pem
        - standards_20260315: Auth-related functions need @auth_required decorator

        Based on existing implementation, here's the refresh token design...

You: Services can't communicate after Docker Compose starts

Claude: [Cross-project query] Found similar issue in project web-dashboard:
        - fix_20260310: Docker network communication failure
        - Solution: Use custom network network_mode: bridge
        - note_20260310: Need to explicitly declare networks in docker-compose.yml

        Generated same configuration for new project...

You: Implement user order query endpoint

Claude: [Query standards] According to project standards:
        - All endpoints return wrapped Result<T> format
        - Pagination uses PageRequest(page, size)
        - Async functions must include _async suffix

        Generated code:

        async def get_orders_async(req: PageRequest) -> Result[List[Order]]:
            ...

/var/lib/project_memory_ai/
├── _metadata.json              # Global metadata (project list, statistics)
├── _stats.json                 # API call statistics
│
├── project_a/                  # ProjectA directory (project name)
│   ├── _project.json           # Project metadata (id, name, info, _version)
│   ├── _tags.json              # Tag registry (tag -> description, aliases)
│   ├── _group_configs.json     # Group configurations (optional)
│   │
│   ├── features/               # Features group
│   │   ├── _index.json         # Feature index (id -> summary, status, tags...)
│   │   ├── feat_20260408_001.json  # Feature details
│   │   └── feat_20260408_002.json
│   │
│   ├── fixes/                  # Bug fixes group (same structure as features)
│   ├── notes/                  # Notes group (same structure as features)
│   └── standards/              # Standards group (same structure as features)
│
├── project_b/                  # ProjectB directory
│
└── .archived/                  # Archived projects
    ├── 20260408_123456_project_a.tar.gz
    └── 20260408_123456_project_a.meta.json

# Use test script
./scripts/run_tests.sh

# Or manually run
pytest test/ -v --cov=src/business --cov-report=html