Otto

Otto is not building travel inventory or a servicing desk. It is an AI front end on other people's infrastructure, and knowing which parts are which tells you what to expect.

The booking rails are Spotnana's. Spotnana's own announcement states that "Otto integrates with Spotnana to combine NDC fares, GDS inventory, and direct supplier APIs," and that "servicing is provided through a seamless handoff to a Direct Travel counselor." Otto's launch post adds Booking.com to the hotel mix. So the fares, the change and cancellation rules, and the person on the phone all come from established corporate-travel plumbing — the same platform that powers Direct Travel itself.

What Otto itself builds is the reasoning layer: parsing an instruction, reading your calendar, remembering preferences, assembling an itinerary, and executing the transaction without redirecting you to a supplier site. As of the 2024 seed round the founder described the approach as "fine tuning off-the-shelf AI models using their own data" rather than training from scratch.

The practical consequence: judge Otto on the quality of the assistant, not on inventory breadth or after-hours support, both of which it inherits. It also means the usual OTA caveat applies — the ticket you end up holding is governed by the airline's or hotel's rules, not Otto's.

This is the part that makes Otto interesting to anyone who works inside an assistant rather than a browser, and it holds up to inspection.

We probed the endpoint on 14 August 2026. https://api.ottotheagent.com/mcp is a real, spec-compliant remote MCP server over Streamable HTTP. An unauthenticated call returns 401 with a proper WWW-Authenticate: Bearer resource_metadata=... header. The protected-resource document names it "Otto Travel MCP" and declares two scopes, mcp:read and mcp:write. The authorization-server metadata advertises dynamic client registration, PKCE with S256, refresh tokens and the OAuth device-code grant. That is a more complete OAuth implementation than many vendors ship, and it means any MCP client can connect, not just the two with directory listings.

Otto is listed as "Otto Travel" in both the Claude and ChatGPT connector directories, with setup instructions on the vendor's site. Any other client can add the URL as a custom remote server.

Worth understanding before you connect it: mcp:write is what lets the assistant complete a booking. Granting it — and Otto's own Claude instructions suggest clicking "Always allow" — means a language model can spend money on your card without a further confirmation step in the client. That is the point of the product, and it is also the risk of the product. If you want to try it without that exposure, connect it and keep approvals manual for the first few trips.

Otto is free, with real specifics attached: "Free – no credit card required. No contracts, no agent-assist fees, no minimum spend," under a headline of "Try Otto free for 1 year." There is no published price list of any kind.

The revenue model is commissions on bookings. GeekWire reported flatly at launch that "the startup makes money on commissions," and in 2024 the founder told TechCrunch that model costs "will be far lower than what the company will earn from affiliate revenue per booking." This is how most online travel agencies work, and it is the reason the free year can exist.

Two things a buyer should hold in mind rather than worry about. First, the end of the free period is not established. GeekWire described it in December 2025 as free "for the next 12 months," which reads as a window closing around December 2026; the site today says "free for 1 year," which reads as per-signup. Nothing published says which it is or what follows, so if you are standardising a team on Otto, ask before you migrate.

Second, because the revenue is per booking, ranking is worth a question. Otto's terms say only that it "may consider several factors when determining the sort order, including without limitation, availability, the User's preferences, distance from applicable location, etc." The clause neither states nor denies that supplier compensation is one of those factors. That is ordinary for the sector and not evidence of anything — but if you are relying on the assistant to pick the fare rather than picking it yourself, it is a fair thing to ask them directly.

Letting an AI transact raises a question most software listings never have to answer, so it is worth reading Otto's terms on the point rather than assuming.

Three clauses matter. Otto disclaims recommendation quality: the terms state that recommendations "are based on algorithms and data available at the time of the query and that Otto does not guarantee the accuracy or suitability of any recommendations." Supplier relationships are pushed downstream: "any dealings with Suppliers are solely between the User and the Supplier," with Otto "not responsible for any breach of the Supplier Rules by either the User or the Supplier." And total liability is capped at "the greater of (A) the amount of fees (excluding any amounts paid or payable to any third-party, including supplier(s)) the user has paid to Otto in the twelve (12) months prior... or (B) $100.00."

None of this is unusual — it is close to standard OTA boilerplate, and a paid competitor's cap would often be similarly modest. But the arithmetic is specific here: while the service is free, fees paid to Otto are zero, so the cap is $100 against a product whose whole purpose is to autonomously buy airfare. In practice your recourse for a genuinely bad booking runs through the airline's or hotel's own rules and the Direct Travel counsellor, not through Otto. Buyers who would want contractual remedies on travel spend should treat that as a reason to keep a human in the loop, not necessarily a reason to pass.

Otto Trip, Inc. is based in Seattle (1522 Western Ave, Ste. 24097). CEO Michael Gulmann spent roughly 14 years at Expedia and Egencia, latterly as SVP of Consumer Product and as Egencia's Chief Product Officer; CTO Chundong "CD" Wang led Uber's airport engineering teams and held roles at CloudKitchens, Snap and Microsoft. For a travel-booking product, that is directly relevant experience rather than a generic AI team.

Funding is a $6 million seed announced 22 August 2024, led by Madrona Ventures with participation from Direct Travel, plus angels described by TechCrunch as former C-suite executives from Expedia, Orbitz, Uber and Farecast. We found no publicly announced round since; that is an absence of evidence rather than a finding, but it means the disclosed capital behind Otto is $6M raised two years ago.

The partner relationships are worth laying out plainly, because they are not arm's-length. Steve Singh founded Concur, is a managing director at Madrona, is executive chairman of Direct Travel, and TechCrunch reported he would "assume a similar position on Otto's board"; GeekWire reports he is also executive chairman at Spotnana. So Otto's booking platform, its human-servicing partner, its lead investor and its chairman all sit in one connected group.

Read that both ways. It is very likely why a young startup has NDC and GDS content and 24/7 counsellor coverage at all — access that would normally take years and far more capital. It also means those partner choices were made inside a family of related companies, so a buyer should not read "built on Spotnana, serviced by Direct Travel" as three independent parties vouching for each other.

Otto is young and shipping, and the honest signals are small.

The vendor's own launch post is unusually candid about scale: "After nine months in a closed beta with hundreds of users and more than 100 completed bookings, Otto is now available to the public" (4 December 2025). Flights and hotels have been live since early 2025; car rental was added 14 July 2026 and is included at no extra cost.

Independent adoption signals are consistent with that. The iOS app averages 4.6 from 10 US App Store ratings — too few to say anything about quality, but a fair indication of user base. Google Play shows 500+ downloads. Both apps were updated on 8 August 2026, six days before this listing, so development is clearly active. We found no G2, Capterra or Product Hunt review corpus; the pages that surface in search under "Otto review" are directory aggregators recycling the vendor's own copy, not hands-on testing.

On the company-controls side, Otto's launch post says: "Through a partnership with Direct Travel, we are piloting programs with companies that want flexibility for travelers and control for administrators," and GeekWire reported these as pilots "with two companies Gulmann could not yet name." If you need policy enforcement, approval flows or admin reporting, treat that as a conversation to have with sales rather than a shipped, self-serve feature — unless something has changed since December 2025, which we could not confirm.

A narrow but checkable point for anyone who has to clear tools through a review.

Otto's privacy policy is dated 15 April 2025 — before the public launch, the MCP connectors and car rental. It contains an "Artificial Intelligence" section listing purposes including sorting results, conversational answering, personalisation, embeddings and text generation. What it does not do is name any subprocessor: it has zero mentions of Spotnana, Direct Travel, or any LLM provider. It also has zero mentions of "calendar," despite calendar access being a headline feature that reads your meetings.

We found no statement either way on whether customer data trains models; the only "training" references concern recordings kept for quality-control purposes. So on model training, subprocessors, and calendar-data handling, the position is not established from public documents. For an individual traveller that is likely immaterial; for anyone doing a vendor security review, those are the three questions to send them.

It fits well if you are a frequent business traveller or a small team with no travel desk, you already live inside Claude, ChatGPT, Slack or Teams, and you want booking to happen in that conversation rather than in eleven browser tabs. The free year makes trying it genuinely low-cost, and there is a human counsellor behind it when something goes wrong at 2am — which is more than most AI travel tools can say.

It fits poorly if you need enforced travel policy, approval chains or spend reporting today; if you need contractual remedies on travel spend; or if you are uncomfortable granting an LLM write access to a payment method. It is also not the right tool if you already have a TMC — Otto's pitch is explicitly aimed at the people who don't.

The honest way to evaluate it: it costs nothing to run three real trips through it and compare the fares and the itineraries against what you would have booked yourself. That is a better test than any listing, including this one.