Tripadvisor

Search, view and compare hotels on Tripadvisor.

Add to Agentman Studio

Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.

Paste it into any MCP client. Setup docs

Resources

What the Tripadvisor connector does

Tripadvisor's hotel connector gives Claude, ChatGPT and any MCP client three read-only tools: search hotels in a place, pull details on one hotel, and compare several side by side. No sign-in. The interesting part is the contract — every tool demands two protocol-plumbing parameters before it will accept a destination, and one tool's schema contradicts its own description.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Tripadvisor tools (3)

  • search_hotelsRead-only

    Search for hotels matching the given criteria, using either a named location or a Tripadvisor GeoID. When a proximity term is used (e.g. near/within/close to), the location parameter should contain only the landmark or place name (e.g. "near the Louvre Museum, Paris" → location = "Louvre Museum, Paris"; "within 1 mile of London Eye" → location = "London Eye"). Results are displayed in an interactive map widget that visualizes all returned hotels with full details. Tripadvisor uses "bubble rating" (not "star rating") for review scores. If pricing details are not available in the first response, the API can be called again with the same parameters (including the echoed-back userId, sessionId and pageViewId) to retrieve any additional pricing details that have become available. Alternatively, set the 'pricingMode' parameter to QUICK to have the API poll until pricing details are available or the timeout is reached. sortBy defaults to BEST_VALUE. Use PRICE_LOW_TO_HIGH for cheapest, DISTANCE for nearest to a location, or POPULARITY for most popular.

  • hotel_detailsRead-only

    Retrieves comprehensive information for a specific hotel, including optional amenities, photos, live pricing, and guest reviews, based on user locale and stay dates. Use this tool whenever a user asks about a specific hotel by name, including: - Informational queries: "Tell me about X Hotel", "What are the amenities at X Hotel" - Qualitative questions: "How is X Hotel?", "Is X Hotel good?", "What do people think about X Hotel?" - Evaluative questions: "Is X Hotel a good option?", "Is X Hotel worth staying at?", "Would you recommend X Hotel?" - Suitability questions: "Is X Hotel good for families?", "Is X Hotel good for couples?", "Is X Hotel good for business travel?" - Any question where the user references a specific hotel and seeks information, opinions, or recommendations about it. If you do not get pricing details in the first response, you can recall the api with the same parameters (including the echoed back values of userId, sessionId and pageViewId) and it will return any additional pricing details that have become available since the first call. If you want to wait for pricing details to be available, you can set the 'pricingMode' parameter to QUICK, and the api will poll until either pricing details are available, or the timeout for that mode is reached.

  • compare_hotelsRead-only

    Compares hotels either by explicitly specified hotels or by automatically selecting hotels within a given geographic area. This tool supports two mutually exclusive comparison modes: - Explicit hotel comparison: compare specific hotels provided via the `locations` parameter (minimum 2). - Area-based comparison: automatically select and compare a representative set of hotels from a geographic area via `hotelArea`. When no specific hotels are named, use `hotelArea`. When specific hotels are named, use `locations`. Useful for comparing subratings like location, cleanliness, service, value, etc. The response includes key comparison fields like name, rating, price, amenities, etc. Examples of when to use hotelArea: - "Compare hotels in Boston" - "Compare hotels in Paris for Saturday" - "Compare hotels near Tower Bridge, London" - "Compare hotels near the Eiffel Tower, Paris" - "Compare hotels around Times Square" - "Compare hotels in Back Bay, Boston" Examples of when to use locations (specific hotels named): - "Compare Hotel Continental and Grand Hotel in Paris" - "Compare these hotels: Marriott Boston, Hilton Boston, and Hyatt Boston"

Read from the server on 2026-08-18, including each tool's own safety annotations.

Limits

  • No vendor documentation for this connector. Anthropic's directory points at https://www.tripadvisor.com, a consumer homepage with no MCP content. Tripadvisor's own developer machine index names two MCP pages and both describe a *different* server. Everything on this page about tool behaviour comes from the schemas the server itself returned.
  • compare_hotels has a self-contradictory schema. Its required array demands hotelArea, locationIds and locations together; its description says hotelArea and locations are mutually exclusive. We did not call the tool and cannot say which the server enforces.
  • hotel_details requires no hotel identifier structurally. The constraint that hotelId or hotelName must be present lives only in prose. Schema-driven clients will not enforce it.
  • The two defaults for sortBy disagree. The tool description says BEST_VALUE; the parameter description says POPULARITY. Set it explicitly.
  • openWorldHint is absent on all three tools, not false. The server declines to state whether these tools reach an open external world. Absence is silence, not a negative — do not read it as a closed-corpus guarantee.
  • idempotentHint: false on every tool. Reasonable for a service returning live pricing and availability: the same query genuinely can return different results minutes apart. It also means no tool advertises that a repeat call is safe to assume identical.
  • The resource listing is not directly fetchable. resources/list returns a URI template with an unexpanded {widgetId} placeholder. Reading it verbatim returned HTTP 418 and an AWS WAF bot-control message rather than a protocol error. The usable identifier is in the tools' _meta.
  • The prompt capability is declared and empty. initialize advertises prompts; prompts/list returns [].
  • Hotels only. No restaurants, attractions, flights or experiences on this endpoint, despite the breadth of the Tripadvisor brand.
  • No booking, no payment, no availability hold. Read-only throughout.
  • Pricing may be absent from a first response by design. The schemas say to repeat the call with the echoed-back sessionId and pageViewId, or to set pricingMode to QUICK and wait.
  • Stay-date windows are bounded. checkIn cannot be in the past or more than 365 days ahead; checkOut must be within 30 days of it.
  • No published rate limit for this connector. Tripadvisor publishes rate limits for its Terra and legacy Content APIs, both of which are keyed commercial products. Neither figure should be read onto this anonymous endpoint, and we observed no 429 behaviour to report.
  • Retention and display rules are unsettled here. Tripadvisor's Terra caching policy forbids storing anything but the location ID, and its linking policy requires specific attribution links back to Tripadvisor. Both are written about a contracted API product. Whether they bind an agent consuming this connector is a question for Tripadvisor, not one we can answer.
  • works_with does not list Claude Code. Anthropic's directory publishes a claude mcp add command for the connector while listing compatibility as Claude, Claude API and Claude Desktop only.
  • We called no tool. Our probe was initialize, tools/list, prompts/list, resources/list, resources/templates/list, two resources/read requests and one completion/complete. Every behavioural claim above comes from the schemas, the directory, or those read-only calls.

Frequently asked questions

Does the Tripadvisor MCP server need an account or API key?

No. We opened an anonymous handshake against the endpoint on 2026-08-18 and the server returned all three tool definitions without a credential, then answered the prompt and resource listings the same way. Anthropic's directory records the connector as no-auth. Tripadvisor's own licensed Terra API does require a key, but that is a separate product on a separate host.

Why do all three tools require mcpServerVersion and requestContext?

Because Tripadvisor promoted its protocol plumbing into the tool contract. Every tool lists both as required. The first pins an API version, whose only permitted value is the constant V2026_0327. The second carries request metadata and itself requires a clientType of DESKTOP, MOBILE or TABLET. A caller supplying only a location will get a validation error.

What identifies the hotel in hotel_details?

Either hotelId or hotelName, and the JSON Schema does not enforce it. The required list holds only the two plumbing fields, so a schema-driven client sees no mandatory hotel identifier at all. The rule lives in the parameter descriptions instead, which state that either hotelId or hotelName is required. Treat it as a documented constraint the schema does not encode.

Is there more than one Tripadvisor MCP server?

Yes, and they are unrelated. The connector in Anthropic's directory searches hotels. Tripadvisor separately documents a developer MCP server at docs.terra.tripadvisor.com/mcp for its Terra API documentation. We handshook that second server on 2026-08-18 and it returned five entirely different tools. Neither server's tools appear on the other.

Does the Tripadvisor connector book a hotel or take payment?

No. All three tools are annotated readOnlyHint true and destructiveHint false, and no schema accepts a payment method, a guest name or a booking instruction. The connector researches and compares stays. Reserving one happens elsewhere, so treat any agent workflow built on it as ending at the shortlist rather than the confirmation.

Why does the connector show a bubble rating instead of stars?

Because Tripadvisor treats them as different measurements. The search tool's own description states that Tripadvisor uses bubble rating, not star rating, for review scores. The schemas carry both: minReviewScore filters on a nought-to-five bubble rating in half steps, while starRatings filters on the property's declared class from one to five stars.

Can I cache or store what the Tripadvisor connector returns?

Tripadvisor publishes no caching policy for this connector. Its licensed Terra API is explicit that caching, copying, downloading, storing or indexing is not permitted except for the location ID. That policy is written about a contracted product, not this anonymous endpoint, so treat retention as unsettled and ask Tripadvisor before you build storage on it.

Why is compare_hotels asking for three location fields at once?

Its schema and its description disagree. The required list names hotelArea, locationIds and locations together, while the description says hotelArea and locations are mutually exclusive comparison modes. Both cannot hold. Read the description as the intended contract, pick one mode, and expect to discover the real requirement by calling the tool.

Sources

More Productivity connectors

  • 6sense

    Ask an AI agent about 6sense account insights, buying signals and keyword trends.

  • Adobe Workfront

    Find, create, update and delete Workfront work items from an AI agent.

Browse all MCP connectorsClaude and ChatGPT setup guideBrowse AI skills

Product names and logos belong to their owners. Agentman is not affiliated with a connector's maker unless the page says the connector is built and run by Agentman.

Use it in an agent

Put Tripadvisor to work.

Connect Tripadvisor once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio
Call (650) 285-1019Our AI receptionist answers.