Maître d'MCP

Maître d'MCP is an MCP-based system that helps you discover restaurants, check availability across multiple booking platforms, and place reservations through natural conversation with chat agents integrated into the MCP framework. It combines Google Places discovery, weather context, dining companion handling, and cross-platform booking to streamline planning, scheduling, and follow-up for your dining experiences.

How to use

You interact with Claude-enabled chat agents to search for venues, refine by preferences, and book reservations across supported platforms. Start by setting up your profile with home and work locations, preferred cuisines, and any dietary restrictions. Then you can ask for nearby restaurant suggestions, ask for highly rated options, or request group-friendly venues. When you choose a restaurant and time, the system checks availability on both Resy and OpenTable, handles any dietary considerations for your dining companions, and can add the reservation to your Google Calendar with a single click.

How to install

Prerequisites: You need Python 3.11+ and Claude Desktop installed. You also require a Google Cloud Places API key. Optional but recommended is an OpenWeather API key for weather-aware suggestions.

# 1) Clone the project
git clone <repo>
cd restaurant-mcp

# 2) Create and activate a virtual environment
python3 -m venv .venv
source .venv/bin/activate

# 3) Install dependencies in editable mode
pip install -e ".[dev]"

# 4) Install browser automation support for Playwright
playwright install chromium

Run the interactive setup to securely store secrets and generate your Claude Desktop config. This step encrypts credentials and produces a ready-to-use Claude Desktop configuration that references a master key for your MCP.

source .venv/bin/activate
python -m src.setup

During setup you will be prompted for:

• Google API Key (required) — Google Places API key
• OpenWeather API Key (optional)
• Resy credentials (optional) — for automated Resy booking
• OpenTable session (optional) — CSRF token and browser cookies for OpenTable booking

The script outputs a Claude Desktop config containing the RESTAURANT_MCP_KEY you will paste into the config file.

Configure Claude Desktop to load your MCP configuration. The config file should reference the generated master key and point to the Python runtime that executes the MCP. After configuring, restart Claude Desktop to enable the restaurant tools.

## Additional content

Security and deployment considerations include encryption at rest for all credentials, strict file permissions (credentials directory: 0700, files: 0600), and a resilient runtime with retries, circuit breakers, and a three-layer booking fallback to handle API or network issues without user disruption.

Important setup details include how to extract and provide OpenTable session values if you need to use automated booking against OpenTable. If your cookies expire, re-run the setup to refresh the session values.

To monitor usage costs and API activity, you can query the built-in api\_costs tool to view costs and cache statistics.

## Tools and capabilities overview

The MCP exposes a collection of tools to manage preferences, groups, and bookings, as well as to discover and reserve venues. The tools cover profile setup, preference management, companion handling, group management, blacklist handling, search, availability checks, reservations, cancellations, visit logging, recommendations, history, and cost tracking.

## Security model and best practices

Your secrets are stored securely using a master key. Sensitive credentials are encrypted at rest, and access is restricted with file permissions. If you are using legacy modes, the system can fall back to older methods, but the recommended approach is to store secrets in encrypted form and load them only at runtime.

## Troubleshooting and notes

If you encounter OpenTable bot protection prompts or authentication errors, ensure you provide valid session cookies and CSRF tokens during setup. If you see API rate or reliability issues, the system includes retry with exponential backoff and a per-service circuit breaker to minimize impact.

Always verify your API keys and credentials are up to date. When adding new group members or updating dietary restrictions, re-run the profile setup to ensure these changes propagate to future searches and bookings.

## Available tools

### setup\_preferences

First-run profile setup including home, work, cuisines, and dietary preferences

### get\_my\_preferences

View current user preferences

### update\_preferences

Change specific user preferences

### manage\_person

Add, update, or remove dining companions

### list\_people

Show all saved companions

### manage\_group

Create, update, or remove groups of companions

### list\_groups

Show all saved groups

### manage\_blacklist

Block or unblock specific restaurants from future suggestions

### search\_restaurants

Find restaurants by cuisine, location, or rating criteria

### check\_availability

Check time slots across Resy and OpenTable

### make\_reservation

Book a table and generate a calendar link

### cancel\_reservation

Cancel an existing reservation

### my\_reservations

View upcoming reservations

### store\_resy\_credentials

Save Resy login data securely (encrypted)

### store\_opentable\_credentials

Save OpenTable login data securely (encrypted)

### log\_visit

Record a restaurant visit in your history

### rate\_visit

Rate a past visit and add notes

### visit\_history

View dining history and past experiences

### get\_recommendations

Get personalized restaurant recommendations

### search\_for\_group

Find venues suitable for a group with merged dietary needs

### api\_costs

View API usage costs and cache statistics