DoorDash MCP Server

MCP server that wraps DoorDash's Drive API. Lets AI agents (Claude, Cursor, etc.) search restaurants, place delivery orders, and track them in real time.

Comes with built-in mock data so you can try it without DoorDash API credentials.

Architecture

graph TB
    subgraph "MCP Clients"
        C1[Claude Desktop]
        C2[Cursor]
        C3[Claude Code]
    end
    
    subgraph "MCP Server"
        T[Transport - stdio / SSE]
        S[Server Core]
        
        subgraph "Tools"
            RT[Restaurant Tools]
            DT[Delivery Tools]
            TT[Tracking Tools]
            AT[Account Tools]
        end
        
        subgraph "Service Layer"
            MC[Mock Client]
            DC[DoorDash Client]
        end
    end
    
    DD[DoorDash Drive API]
    
    C1 & C2 & C3 --> T
    T --> S
    S --> RT & DT & TT & AT
    RT & DT & TT & AT --> MC
    RT & DT & TT & AT -.->|live mode| DC
    DC --> DD

Quick Start

No API keys needed — mock mode is the default:

git clone https://github.com/chaituredd/doordash-mcp-server.git
cd doordash-mcp-server
npm install && npm run build

The build step prints the MCP client config JSON. Paste it into your Claude Desktop or Cursor settings.

Or with Docker:

docker compose -f docker/docker-compose.yml up

Tools

12 tools across 4 domains:

Configuration

Copy .env.example to .env:

MCP Client Config

Claude Desktop / Claude Code:

{
  "mcpServers": {
    "doordash": {
      "command": "node",
      "args": ["/absolute/path/to/doordash-mcp-server/build/index.js"],
      "env": {
        "DOORDASH_API_MODE": "mock"
      }
    }
  }
}

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "doordash": {
      "command": "node",
      "args": ["/absolute/path/to/doordash-mcp-server/build/index.js"]
    }
  }
}

Development

npm install          # install deps
npm run dev          # watch mode
npm test             # run tests
npm run test:watch   # tests in watch mode
npm run typecheck    # type checking

Project Layout

src/
├── index.ts          # entry point, transport setup
├── server.ts         # tool/resource/prompt registration
├── config.ts         # env validation (Zod)
├── tools/            # one file per tool domain
├── services/         # DoorDash clients (real + mock)
├── data/             # mock restaurants, menus, deliveries
├── types/
└── utils/            # logger, errors, validation

Live Mode

1. Sign up at developer.doordash.com (sandbox is free)
2. Grab your credentials and run:

DOORDASH_API_MODE=live \
DOORDASH_DEVELOPER_ID=your_id \
DOORDASH_KEY_ID=your_key \
DOORDASH_SIGNING_SECRET=your_secret \
npm start

New accounts start in sandbox. Production requires a separate application to DoorDash.

How It Works

Implements the Model Context Protocol spec. AI agents call tools through MCP, the server translates those into DoorDash Drive API requests (or returns mock data), and formats responses for the agent to show the user.

The mock layer is stateful — create a delivery and it'll progress through created → confirmed → enroute_to_pickup → picked_up → enroute_to_dropoff → delivered each time you check status. Makes demos actually useful.

Tech

TypeScript (strict), Node 18+, @modelcontextprotocol/sdk v1.x, Zod for validation, Axios for HTTP, Vitest for tests, GitHub Actions CI, Docker multi-stage build.

Why This Exists

Wanted to learn MCP properly by building something non-trivial with it. DoorDash's Drive API seemed like a good fit since it has a real delivery lifecycle to model. The mock layer turned out to be the most useful part — you can demo the whole flow without any API keys.

License

MIT