Beli MCP Server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Beli MCP Server search for ramen spots in Tokyo and bookmark the top result"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Beli MCP Server
Unofficial Beli MCP server: let Claude and other AI assistants search restaurants, manage bookmarks and read guides on your Beli account.
Overview
beli-mcp is a Model Context Protocol server for the Beli restaurant ranking app. It wraps the unofficial Beli API documented in beli-api-spec, so an AI assistant can search restaurants, bookmark places and read your guides. It signs in with your own Beli account and spaces requests to stay under Beli's rate limit.
Related MCP server: 402 MCP Restaurant Finder
Features
Search - find restaurants in a city and get their Beli business ids
Bookmarks - list bookmarks, check a place and add new ones
Guides - list your guides and read their places
Referral link - fetch your referral link
Read-only escape hatch -
beli_getcalls any GET endpoint from the specRate limit aware - requests are spaced about 3 seconds apart
Tools: whoami, search_restaurants, get_bookmarks, add_bookmark, bookmark_status, list_my_guides, get_my_guide_items, get_guide, referral_link, beli_get. add_bookmark is the only tool that writes.
Tech Stack
Getting Started
Prerequisites
Python 3.10+
A Beli account with an email and password login
Installation
git clone https://github.com/horse-3903/beli-mcp
cd beli-mcp
pip install .Run locally
BELI_EMAIL=you@example.com BELI_PASSWORD=your-password beli-mcpAdd to Claude Desktop or Claude Code
{
"mcpServers": {
"beli": {
"command": "beli-mcp",
"env": { "BELI_EMAIL": "you@example.com", "BELI_PASSWORD": "your-password" }
}
}
}Project Structure
beli-mcp/
├── beli_mcp/
│ ├── __init__.py
│ └── server.py # MCP tools and Beli API client
├── pyproject.toml
├── AGENTS.md # instructions for AI coding agents
└── README.mdEnvironment Variables
Variable | Default | Description |
| none | Beli account email |
| none | Beli account password |
| none | Optional refresh token, used instead of the password while valid (7 days) |
FAQ
Is there an official Beli MCP server? No. This one is unofficial and uses the reverse engineered Beli API.
Does it work with Claude? Yes, with any MCP client such as Claude Desktop, Claude Code or Cursor.
Disclaimer
Unofficial and not affiliated with Beli. Use at your own risk. Calling the API outside the official app may violate Beli's terms and can lead to rate limiting, a suspended or banned account, or loss of data. The authors accept no responsibility for any consequences to your account. Use only your own account, and never share your credentials or tokens.
Available Tools
10 toolsadd_bookmarkB
Bookmark a restaurant by its Beli business id. Idempotent. This writes to the user's account.
Inputs: business_id (required).
beli_getA
Read-only escape hatch: GET any /api/ path from the Beli spec (see beli-api-spec). Use {user} in the path for your user id.
Inputs: path (required), params.
bookmark_statusC
Check whether a restaurant is bookmarked.
Inputs: business_id (required).
get_bookmarksC
List the signed-in user's bookmarks. Category RES is restaurants.
Inputs: category.
get_guideB
Read one guide in full by id.
Inputs: guide_id (required).
get_my_guide_itemsA
List the places in all of the signed-in user's guides (each item names its guide id).
Inputs: no named parameters.
list_my_guidesA
List the signed-in user's Beli guides with titles, audience and place counts.
Inputs: no named parameters.
referral_linkA
Get the signed-in user's referral link.
Inputs: no named parameters.
search_restaurantsA
Search Beli for restaurants. Returns candidates with a business id and address. Check the address, since top hits can be in the wrong city.
Inputs: term (required), city.
whoamiA
Return the signed-in Beli user (id, username, name).
Inputs: no named parameters.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 10 tools
Most tools target distinct resources (guides vs restaurants vs bookmarks vs user), and bookmark_status vs get_bookmarks are separated by singular-check vs list-all intent. Minor overlap between list_my_guides/get_my_guide_items/get_guide and the beli_get catch-all creates some selection ambiguity, but descriptions differentiate them adequately.
A mix of conventions: verb-prefixed actions (list_my_guides, get_guide, search_restaurants, add_bookmark, get_bookmarks) coexist with noun-only tools (bookmark_status, referral_link), the special whoami, and the server-prefixed beli_get. Readable but not a predictable pattern.
10 tools is well-scoped for a domain-specific integration covering users, guides, restaurants, bookmarks, and referrals. Each tool has a clear place without obvious padding.
The surface is read-heavy: it adds bookmarks but offers no remove_bookmark, no guide creation/editing, and no rating/ranking tool despite Beli's core function being restaurant ratings. The beli_get escape hatch mitigates gaps, but common lifecycle operations remain missing.
Maintenance
Related MCP Connectors
Lets Claude, ChatGPT, and Muse read and update your beckett tasks, notes, calendar, and lists.
OpenTable MCP server for ChatGPT, Claude and any AI agent: search restaurants with open tables, check availability, book a reservation, modify a reservation, cancel a reservation, list your reservations, read the reservation book and add guest notes and tags in your own OpenTable account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to OpenTable yourself; Cotlet does not store your password and you need no OpenTable API key. About the official OpenTable API: OpenTable has APIs only for approved partners (checked 27 Oct 2026). Booking, Sync, CRM, Directory and Reviews APIs need an application through the OpenTable partner program, and they work per restaurant, not on a diner account. Setup and FAQ: https://cotlet.dev/integrations/opentable-api-mcp-server
Resy MCP server for ChatGPT, Claude and any AI agent: search restaurants, check availability, book a reservation, list reservations, change a reservation, cancel a reservation, set a Notify alert and remove a Notify alert in your own Resy account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Resy yourself; Cotlet does not store your password and you need no Resy API key. About the official Resy API: Resy has no public API (checked 27 Oct 2026). Its official Resy app runs only inside Claude and ChatGPT: in Claude it shows live availability and sends you to Resy to book, in ChatGPT it can also book a table. Setup and FAQ: https://cotlet.dev/integrations/resy-api-mcp-server
Discover, read and book verified real-world businesses through one endpoint.
Related MCP Servers
- FlicenseDqualityDmaintenanceEnables interaction with a restaurant FastAPI backend for menu management, order processing, customer management, and analytics through natural language.241-
- FlicenseNot gradedqualityDmaintenanceEnables AI agents to browse restaurants and menus for free, and place orders with USDC payment via x402.-
- AlicenseDqualityBmaintenanceProvides complete access to Mealie's API for managing recipes, meal plans, shopping lists, households, and more via natural language from AI assistants.28321MIT
- FlicenseNot gradedqualityDmaintenanceEnables Claude to read and interact with BILLZ 2.0 data such as products, sales, customers, and categories via the BILLZ REST API.-