# opentable [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/markswendsen-code/mcp-opentable  
**GitHub Stars:** 1  
**npm Downloads (last month):** 379  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/opentable

## Description
MCP server for OpenTable - let AI agents search restaurants and make reservations

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "opentable": {
    "command": "npx",
    "args": ["-y","@striderlabs/mcp-opentable"],
    "env": {
      "OPENTABLE_EMAIL": "",
      "OPENTABLE_PASSWORD": ""
    }
  }
}
```

**Requires environment variables:** `OPENTABLE_EMAIL`, `OPENTABLE_PASSWORD` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What the opentable MCP server does

The opentable MCP server gives an AI agent access to OpenTable restaurant search and reservation workflows. It can find restaurants using location, cuisine, rating, price range, dining style, date, time, and party size filters. After identifying a venue, the agent can retrieve restaurant details, check available times, create a reservation, and retrieve the user's existing reservations.

The booking flow supports party size, date, time, and optional special requests. Reservations use a confirmation step rather than being described as an unattended background process. The server is intended for agents that need to handle both discovery and reservation management through OpenTable.

## How it works

The opentable MCP server uses Playwright browser automation to interact with OpenTable. An agent sends structured tool requests, and the connector performs the corresponding OpenTable actions in a browser session.

Account sessions are designed to persist between restarts. The project describes encrypted, per-user credential and session storage, automatic OAuth token refresh, and MFA handling through SMS or email. It also describes browser profiles, rate limiting, UI-change detection, fallback paths, and monitoring as part of the connector architecture.

## Setup and configuration

The package can be installed with npm or started directly with npx:

```bash
npm install @striderlabs/mcp-opentable
npx @striderlabs/mcp-opentable
```

For Claude Desktop, add an `mcpServers` entry that runs `npx -y @striderlabs/mcp-opentable`, then restart Claude. A self-hosted setup is also documented: clone the repository, run `npm install`, and start it with `npm start`; the example says the agent can then connect to `localhost:3000`.

`OPENTABLE_EMAIL` and `OPENTABLE_PASSWORD` can be supplied to use a specific OpenTable account. They are presented as optional configuration variables, and the README recommends keeping the password in a `.env` file rather than exposing it directly in a command.

## Tools and capabilities

The documented agent operations are:

- `search_restaurants` for filtered restaurant discovery.
- `get_restaurant_details` for information about a selected restaurant.
- `check_availability` for checking times against party size and date.
- `make_reservation` for booking a selected time, including optional special requests.
- `get_my_reservations` for viewing the account's reservations.

Search filters include cuisine, location, price range, date, party size, time, and rating. The feature list also mentions filtering by dining style, persistent sessions, automatic MFA handling, and per-user encrypted session storage.

## Limitations and notes

The opentable MCP server depends on browser automation and OpenTable's web interface, not a comprehensive public reservation API. Changes to that interface can affect the connector, although the project says it detects UI changes and updates connectors when needed. OpenTable account access is required for account-specific actions such as making or viewing reservations.

The repository identifies the project as MIT licensed. Its compatibility table lists Node.js 18+, MCP SDK 1.0.0 or newer, Claude Desktop, Claude API with Claude 3.5 Sonnet or newer, and Anthropic SDK 0.20 or newer as tested components. The README also reports an 85% or higher task completion rate, but this is a project-reported metric rather than a guarantee for every reservation attempt.

_Full upstream README: https://allmcps.com/mcp/opentable/readme_