mcp-server

# Druling MCP Server ## Install all dependencies - Run `pip install -r requirements-dev.txt` ## How to run app using Docker with PostgreSQL - Install Docker Desktop - Run `docker compose up --build` - Run `docker compose down` to stop all services ## How to run locally without postgres or docker - In database/core.py change the DATABASE_URL to sqlite - Run `make run` ## How to run tests - Run `pytest` to run all tests --- ## How to Add a New MCP Server MCP (Model Context Protocol) servers provide tools and prompts for AI integrations. Follow these steps to add a new MCP server: ### Step 1: Create the MCP Server Directory Create a new directory under `src/servers/` for your MCP server: ``` src/servers/<provider>/<service>/ __init__.py mcp.py outputs.py prompts.py ``` For example, to add a new "calendar" service under Google: ``` src/servers/google/calendar/ __init__.py mcp.py outputs.py prompts.py ``` ### Step 2: Define Output Models (`outputs.py`) Create Pydantic models for your tool outputs: ### Step 3: Define Prompts (`prompts.py`) Create prompts for your MCP server (optional): ### Step 4: Create the MCP Server (`mcp.py`) Create the MCP server class by extending `BaseMCPServer`: ### Step 5: Register the MCP Server Update `src/setup/mcp.py` to register your new MCP server: ```python from src.servers.google.calendar.mcp import CalendarMCPServer # Add to existing imports and initializations calendar_server = CalendarMCPServer() MCP_PATH = { # ...existing servers... "calendar": calendar_server, } ``` The server will automatically: - Be mounted at `/<server_name>` (e.g., `/calendar`) - Have session management handled by the lifespan context - Include authentication and context middleware --- ## How to Add a New REST API REST APIs are used for standard HTTP endpoints. Follow these steps to add a new API: ### Step 1: Create the API Directory Create a new directory under `src/app/` for your API: ``` src/app/<feature>/ __init__.py api.py ``` For example, to add a "users" API: ``` src/app/users/ __init__.py api.py ``` ### Step 2: Create the API Router (`api.py`) Create the FastAPI router with your endpoints: ### Step 3: Register the API Router Update `src/setup/api.py` to include your new router: --- ## Project Structure Overview ``` src/ ├── main.py # FastAPI app entry point ├── setup/ │ ├── api.py # REST API route registration │ └── mcp.py # MCP server registration ├── app/ # REST API endpoints │ ├── health_check/ │ ├── auth/ │ └── <your_api>/ ├── servers/ # MCP servers │ ├── google/ │ │ ├── gmail/ │ │ ├── drive/ │ │ └── <your_service>/ │ ├── druling/ │ └── <your_provider>/ ├── core/ │ ├── service/ # Base MCP server class │ ├── middleware/ # Auth and context middleware │ ├── exceptions/ # Custom exceptions │ └── utils/ # Utility functions └── clients/ └── backend/ # Backend API client ``` --- ## Key Concepts ### MCP Server - Extends `BaseMCPServer` from `src.core.service` - Provides tools (functions) and prompts for AI agents - Uses `@self._mcp.tool()` decorator to register tools - Uses `mcp_meta()` for tool metadata - Access user context via `self.get_context()` ### REST API - Uses FastAPI's `APIRouter` - Registered via `app.include_router()` - Standard REST conventions with proper status codes ### Backend Client - Use `BackendClient` from `src.clients.backend.client` to call backend services - Supports `get()`, `post()`, `put()`, `delete()` methods - Pass `context` for authenticated requests