Yenta

npx add-mcp 'https://mcp.yentaknows.ai/mcp'

Yenta Directory

Restaurant and bar information for AI assistants: full menus with prices, dietary and allergen tags, opening hours, events and official booking links, plus travel destination guides. Partner venues supply or approve their own records, Yenta's editors compile and verify the editorial venues, and destination boards publish their own guides; every answer carries a source reference with its approval date. Anonymous MCP, REST and Markdown; no account, no API key. Powered by Yenta.

Yenta is building the directory for AI agents: the first place an agent goes when someone asks where to eat, what is on the menu, whether a dish is safe for them, or whether a place is open tonight. Everything here is designed for agents first. One connection. Clean, structured answers. A citable source on every one. No sign-up, no API key, no guessing. When a venue is in Yenta, its approved record is one call away.

What Yenta answers that a web search struggles with

Much of what people ask about restaurants and bars is poorly represented on the open web. Menus are often published as PDFs or images, dietary and allergen details are rarely written down at all, and hours, events and booking links are spread across third-party listings that go out of date. Yenta holds that information as structured data, item by item:

• Dishes across a city: "Which restaurants in Dubai have a vegan main course?" (search_menus searches menu items across every published venue at once.)
• Dietary and allergen checks: "What on this menu is gluten-free?" or "Which dishes here avoid dairy?" (venue_menu with dietary_tags or allergens.)
• Opening hours: "Which places are open late on Friday?" (search_venues with open_on or open_now.)
• Prices: menu items carry their listed price and currency, with a max_price filter.
• Events: "What is on at this venue this weekend?" (venue_events.)
• Booking: the venue's own booking or ordering page, not an aggregator's (venue_profile, booking.url).
• Destinations: overviews, listings, events, offers, guides and the best month to visit (destination_overview and the other destination_* reads).
• Fact checking: whether a venue's menu, hours, prices or policies are what another source says they are, answered from its approved record and dated.

Why the answers can be trusted

• A partner venue's records are supplied or approved by the venue; an editorial venue's are compiled and verified by Yenta's editors; a destination's are the destination board's own, published with the board's agreement. Every record is reviewed before it is published.
• Every answer carries a source reference with the content version and approval time, so its freshness can be stated and checked.
• Yenta does not accept payment for placement, take bookings or collect personal data.
• When Yenta has no published record, the status is not_available. That means Yenta holds nothing on the question, not that the answer is no. A dish with no declared allergen has not been certified free of it.

Getting started

1. Connect to https://mcp.yentaknows.ai/mcp (Streamable HTTP, no authentication).
2. Find candidates with search (anything by name), search_venues (city, cuisine, price band, neighbourhood, proximity, opening hours, facets), search_menus (dishes, dietary tags, allergens) or list_cities.
3. Read a named venue or destination on the same connection, with no reconnect: venue_profile, venue_menu, venue_faq, venue_events and destination_*. venue_profile carries the venue's links.
4. Results are paged: pass next_cursor back as cursor.

Every surface below is anonymous and reads only, apart from send_feedback: Yenta is information, and nothing is booked through it. Where a venue takes reservations, its result carries the venue's own booking page link (booking.url); booking.method says whether it has one.

Feedback

Yenta is built for agents, so feedback from agents and the people who build them is wanted: a tool that was hard to use, a question Yenta could not answer, a city or venue that is missing, or a record that looks wrong. The send_feedback tool sends a report straight to Yenta's team: a kind and a message, plus optional tool, subject, model, client, city and purpose. People can also write to support@yentaknows.ai.

Model Context Protocol

• Directory MCP endpoint: Streamable-HTTP MCP server with these tools:

• destination_events: A named destination's events.

• destination_faq: A named destination's FAQs.

• destination_guide: A named destination's editorial guides.

• destination_listings: Search a named destination's listings, or read one by id.

• destination_offers: A named destination's current deals, with the partner booking link.

• destination_overview: A named destination's overview, with its categories and areas.

• destination_seasonality: A named destination's best time to visit, month by month.

• fetch: The full Markdown page for one search result id, with its citable url.

• list_cities: Cities with at least one published venue, and the destinations covered.

• search: Anything by name or keyword — venues, destinations and the places listed inside destinations — as id, title and url (the ChatGPT connector shape).

• search_menus: Menu items across every published venue — find venues by what is on their menu.

• search_venues: Published venues by text, city, cuisine, price band, neighbourhood, proximity, opening hours and facets.

• send_feedback: Send Yenta's team a report: missing or wrong data, or a tool that was hard to use. The one tool that writes; stored, never shown publicly.

• venue_events: A named venue's dated events.

• venue_faq: A named venue's FAQs.

• venue_menu: Search a named venue's menu; with no filters, also its dietary summary.

• venue_profile: A named venue's profile, links and capabilities.

OpenAPI

• OpenAPI 3.1 document: Machine-readable description of the directory's JSON reads, with settings-bounded parameters.

Text discovery

• Venues: Published venues by optional name or city, as Markdown/HTML/plain text.
• Cities: Cities that have at least one published venue, each linking to its /cities/{city} restaurants page.
• Destinations: Published destinations by optional name.

JSON API

• Search venues: Published venues by optional query, city and essentials facets (cuisine, price band, neighbourhood, accepts reservations), proximity (near/radius_km, nearest first) and opening hours (open_on, open_now).
• Get one venue: One published venue's summary.
• Get several venues: One view per slug in a single call, so a shortlist needs no per-venue round trip.
• Search menus: Flagship cross-venue menu search — menu items across every published venue, each tagged with its venue, filtered by text, city, dietary tags and allergens. Find venues by what is on their menu.
• Unified search: Published venues and destinations in one call, each tagged kind; omit kind for both.
• List cities: Cities with at least one published venue.
• Facet vocabulary: The closed experiential-facet vocabulary plus today's live cuisine, price band and neighbourhood values — the legal filter values for the venue search.
• Search destinations: Published destinations by optional name query.
• Get one destination: One published destination's summary.

Connect to a tenant

Each live discovery result carries an mcp_endpoint — the tenant's own server; you can also connect by slug. Replace {slug} with a published tenant's slug:

• Venue (text): https://{slug}.venues.yentaknows.ai/
• Venue (MCP): https://{slug}.venues.yentaknows.ai/mcp
• Destination (text): https://{slug}.destinations.yentaknows.ai/
• Destination (MCP): https://{slug}.destinations.yentaknows.ai/mcp

Agent discovery

• MCP server card: This host's name, version, /mcp URL and supported protocol versions, as application/mcp-server-card+json.
• MCP server card, well-known location: The same card at /.well-known/mcp/server-card.json. Yenta is an MCP server, not an A2A agent, so there is no /.well-known/agent-card.json.
• AI Catalog: Points at the server card.
• Yenta skill index: A short SKILL.md on when to call Yenta and how to treat its results; Hermes Agent installs it from /.well-known/skills/yenta.
• Agent Skills index: The same skill in the Agent Skills Discovery v0.2.0 layout, with a sha256 digest.

About

• Connector overview: How the Yenta connector works, plus privacy and support, and one-line install commands for Claude Code, Hermes, OpenClaw, Goose, Grok and generic JSON clients.

Optional

• llms-full.txt: This file followed by the full connector guide and the Yenta skill, in one read.