MCP 服务器

menami

Personalized restaurant recommendations, table bookings, and delivery via MCP, CLI, or API, learning user taste and acting proactively.

README

Your personal food agent. AI-powered restaurant recommendations, table bookings, and delivery — via MCP, CLI, or API.

Why Menami?

Most restaurant APIs give you data. Menami gives you an agent that knows you.

It learns your taste. Not just "Italian" — it knows you love smoky mole, hate loud restaurants, and your partner is allergic to shellfish.
It acts on your behalf. Your anniversary is in 3 weeks? Menami already booked a table at the impossible-to-get restaurant that matches both your taste profiles.
It gets better every meal. Every rating, every feedback, every "I tried that place and loved it" makes the next recommendation sharper.
It works everywhere. Use it from Claude, from the terminal, from your own app, or just text it on WhatsApp.

Menami covers San Francisco, New York City, and Mexico City with 3,000+ restaurants in the knowledge graph.

Quick Start

I want to...	Use this
Use Menami inside Claude Desktop or another AI agent	MCP Server
Use Menami from the terminal	CLI
Build an app on top of Menami	REST API

For AI Agents (MCP)

Menami implements the Model Context Protocol so AI assistants like Claude can call it as a tool.

1. Authenticate

npx @menami/mcp-server connect

This opens a browser OAuth flow and saves credentials to ~/.menami/config.json.

2. Add to Claude Desktop

Edit your Claude Desktop config:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "menami": {
      "command": "npx",
      "args": ["@menami/mcp-server"]
    }
  }
}

Restart Claude Desktop. You'll see Menami tools in the tool panel.

3. Try it

"Find me a good sushi place for a date night in San Francisco. Budget is $80-120 per person."

Claude will call menami_consult_agent and return personalized recommendations from your taste profile.

Available Tools

Tool	Description
`menami_consult_agent`	Personalized restaurant recommendations based on taste, location, and occasion
`menami_book_table`	Check availability and book a table
`menami_place_order`	Place a delivery or pickup order
`menami_submit_feedback`	Rate a dining experience — improves future recommendations
`menami_get_taste_profile`	View your taste profile (cuisines, dietary, price range, favorites)
`menami_get_restaurant`	Get restaurant details (menu, hours, reviews, location)
`menami_manage_occasions`	Manage birthdays, anniversaries, and trips for proactive suggestions

Full schema reference: docs/tools.md

For Developers (CLI)

npx @menami/cli login

Commands

# Get personalized recommendations
menami recommend "best ramen for a cold night"
menami recommend --cuisine japanese --occasion date_night
menami recommend                    # interactive mode

# Search the knowledge graph
menami search --city "San Francisco" --cuisine italian
menami search --city "Mexico City" --query "tacos de guisado"

# Get restaurant details
menami restaurant <id>
menami restaurant <id> --json

# View your taste profile
menami profile
menami profile --format full

# Set up your taste profile (interactive wizard)
menami onboard

# Rate a dining experience
menami feedback <restaurant-id> --rating 4 --text "Amazing tonkotsu, but noisy"

# Book a table
menami book <restaurant-id> --date 2026-04-15 --time 19:30 --party 2

# Manage occasions
menami occasions add birthday
menami occasions add anniversary

Global Flags

Flag	Description
`--json`	Output raw JSON
`--help`	Show help for any command
`--version`	Show version

For Apps (API)

Base URL: https://api.getmenami.com

All endpoints require a Bearer token. See Authentication.

Recommendations

curl -X POST https://api.getmenami.com/v2/recommendations \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"message": "date night Italian restaurant"}'

Search Restaurants

curl "https://api.getmenami.com/v2/restaurants/search?city=sf&cuisine=japanese" \
  -H "Authorization: Bearer $TOKEN"

Book a Table

curl -X POST https://api.getmenami.com/v2/bookings \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "restaurantId": "rst_abc123",
    "datetime": "2026-04-15T19:30:00",
    "partySize": 2,
    "guest": {"name": "Alice", "email": "alice@example.com"}
  }'

Submit Feedback

curl -X POST https://api.getmenami.com/v2/feedback \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"restaurantId": "rst_abc123", "overallRating": 5, "feedbackText": "Perfect omakase"}'

Get Taste Profile

curl https://api.getmenami.com/v2/profile \
  -H "Authorization: Bearer $TOKEN"

Full endpoint reference: docs/api-reference.md

Authentication

Menami uses OAuth 2.0 with PKCE — no client secrets needed. Safe for CLIs, native apps, and AI agents.

1. Generate code_verifier + code_challenge (SHA-256)
2. Open browser → /oauth/authorize?code_challenge=...
3. User authenticates
4. Receive auth code at localhost callback
5. Exchange code + verifier for access_token + refresh_token
6. Use Bearer token for all API calls
7. Refresh when expired via POST /v2/auth/refresh

Tokens are stored at ~/.menami/config.json.

Full details: docs/authentication.md

Configuration

Variable	Default	Description
`MENAMI_API_URL`	`https://api.getmenami.com`	Override API base URL

Repository Structure

menami/
├── packages/
│   ├── mcp-server/        # @menami/mcp-server — for AI agents
│   │   └── src/
│   │       ├── cli.ts     # npx entry point (connect + MCP stdio server)
│   │       ├── tools.ts   # 7 MCP tool definitions with JSON Schema
│   │       ├── types.ts   # TypeScript interfaces
│   │       └── auth.ts    # OAuth PKCE flow
│   └── cli/               # menami — terminal CLI
│       └── src/
│           ├── index.ts   # Commander.js (9 commands)
│           ├── commands/  # One file per command
│           ├── client.ts  # HTTP client with auto token refresh
│           └── config.ts  # ~/.menami/config.json persistence
├── docs/
│   ├── api-reference.md   # Full REST API docs
│   ├── authentication.md  # OAuth PKCE details
│   └── tools.md           # MCP tool schemas
└── assets/
    └── logo.svg