Launching today
Tripmojo syncs your booking emails into one shared trip, then Mojo, your AI travel agent, helps run it. Plan in a group chat where Mojo answers, runs polls and finds places. On travel day Mojo tries to check you in, saves your boarding pass, flags delays and gate changes, and tells you when to leave. On the trip it keeps today's plan current and splits expenses. Free on iOS and Android.













Free
Launch Team / Built With

Render Sandboxes: a computer for every agentIsolated compute, persistent state and a real filesystem
Promoted





Connected my Gmail and it rebuilt my October Delhi–Ranchi trip by itself: both IndiGo flights, the right dates, ₹16K total, even the boarding passes and GST invoices in Docs. Really liked that.
You asked which booking email you read wrong, so here's one. Each of my flights shows two travellers under one PNR: "Yuvraj Tiwary" and "Yuvraj Tiwari". I flew alone and every document says Tiwary, so the trip says "You & 1 other" while the travellers page shows only me.
Saw in Built With that Gemini does the extraction with structured outputs. My guess is the schema keeps the shape right but not the spelling: on one of the documents the model "corrected" Tiwary to the more common Tiwari, and the merge added a new person. Do you check that extracted names actually appear in the source email or PDF? And would auto check-in have used the wrong surname here?
@yuvraj_tiwary Thanks Yuvraj, really useful, and glad the rest of the trip came through right.
You called it. None of your documents say "Tiwari". We checked all of them, and every one says Tiwary. The extraction slipped on the spelling of one, and two things on our side should have caught it but didn't:
1. We don't check that an extracted name actually appears in its source email or PDF.
2. Our step that decides whether a new name is the same person or another traveller wasn't switched on for every account yet, so the near-duplicate got added as a second traveller.
We're fixing both: a name has to appear in your documents, and near-identical names on one booking fold into one person instead of adding a traveller.
For auto-checkin, we would have not changed the names and would have used the name from the booking itself.
@sahilgoel Thanks Sahil, really appreciate such quick verification of docs.
Additionally, my e-ticket lists only me as a passenger, but booking still shows two travellers, so a simple comparison check between number of travellers and the passenger count on the ticket could catch this, without being dependent on spellings.
Also I just thought of an edge case of merge step, imagine 2 real people with almost same name on one PNR, like cousins Ankit Sharma and Ankita Sharma, just one letter difference but completely different people, so if both names are on same ticket then they should stay separate?
@sahilgoel @yuvraj_tiwary Both great points. Comparing against the passenger count on the ticket is a cleaner check than spelling, and we'll add it. If the ticket lists one passenger, the booking can't have two.
And yes, Ankit and Ankita on the same ticket stay separate. Two names that both appear on one ticket are two real people. We only fold near-identical names together when the documents never show both.
Love that Mojo flags delays and gate changes with a leave-by time — that's the part of travel day that always stresses me out. Does it nudge the whole group chat too, or just the person whose booking changed?
@ms_harita_kanuri Both! 🙌
When a flight is delayed, cancelled or retimed, Mojo messages the travellers on that flight directly in their own Mojo chat. That's where the personal stuff goes, like the new leave-by time. Mojo also posts the schedule change in the trip's group chat. That way the friend picking you up, or the people on a different flight, know without anyone having to forward a screenshot.
The idea is that the person affected gets what they need to act on, and the rest of the group gets just enough to adjust their plans.
Congrats on the launch! I was just testing manual flight entry by typing just the flight number “6E 1079” fills in the times for today and tomorrow, but for 22 Nov it says “Schedule not found” every time. So the lookup seems to only see maybe 48h ahead, which makes me guess it’s a live flight-status API rather than a schedules feed. Time zones from the airports and the automatic +1 day handled the red-eye well though. I am just curious about when a booking comes in weeks before the trip via email or Gmail sync, do you go back and verify it once it’s inside that window, or does it stay as parsed?
@md_obaidullah_ansari
Thanks for testing this so closely, and glad the red-eye handling held up!
You read it right. Manual lookup uses a live flight-status feed, and that feed only covers about the next two days, which is why 22 Nov fails. Looking up flights further out is a gap, and we're looking into it.
And yes, we go back and check. A booking imported weeks ahead stays as it was parsed until the flight enters the feed's window. From then on we track it against live data. If the airline moves the time by 15 minutes or more, we update the booking and alert you. Gate changes, delays and baggage belts come through the same way. A schedule-change email from the airline also updates the existing booking rather than adding a new one.
@garg_rish
Thanks, super helpful! Since you’re already tracking flights live, have you ever thought about building a schedule cache from that history? Most flight numbers repeat on the same route and time each week, so you could fill in times beyond 2 days with a “provisional” tag and confirm once the flight is in the window. It would also catch parse errors at import instead of 2 days out. The main risk which I see is the schedule change in late March and October. Would that make a paid schedules feed simpler?
@md_obaidullah_ansari
Great idea, and you've named the real tradeoff. A cache built from flight history would cover the routes our travellers fly often, and a "provisional" tag keeps it honest. It would also catch parse errors right at import, which we like a lot.
The weak spots are the ones you pointed out: the March and October schedule changes, plus rarely flown flights we have no history for. A paid schedules feed handles both, but it costs money for every lookup. We'll probably end up mixing the two: history first, the feed for gaps and the season changes. Thanks for thinking this through with us!
Hi team! I'm curious about how Mojo handles AI-generated trip planning when users have multiple constraints, such as budget, travel dates, preferences and group decisions. Do you use an LLM with tool calling and structured outputs to generate itineraries, and how do you validate that the recommendations and booking information are reliable before presenting them to users?
@aman_mishra35 Great question! Yes, Mojo runs on Gemini with tool calling and structured outputs. The schemas are strict and every field we need is required, so the model can't quietly skip things.
The constraints come from the trip itself: dates, budget and preferences from what your group actually said in chat and what you've already booked. If something isn't clear, Mojo asks instead of guessing.
For reliability, the model never invents a place, price or booking detail. It uses tools to look things up: live flight status, visas, places, hotel and activity availability. Every place maps to a real Google Maps listing or curated dataset we have curated by collaborating with multiple travel enthusiasts and open web. Prices are fetched live from the supplier when you book, never taken from what the model wrote. Booking info comes from your actual confirmation emails, checked against live flight data.
We also keep a set of real-world test cases that every change has to pass before it ships.
Still after all these guardrails, LLM can hallucinate and we try to keep our eval set exhaustive which we run daily.
Thanks for asking! 🙂
Hey, I tried Tripmojo with my real train ticket from IRCTC and it got almost everything right. The route, the 33 hour journey, next day arrival, even the exact price. Only thing that felt wrong was Day 1 showing New Delhi, when I'm still on the train in Assam that day. How does the app decide which city a day belongs to when the trip goes overnight?
@aditya_mittal17 Thanks for testing with a real ticket, and glad the route, timing and price all came through!
You found a real gap. On a travel day, we label the day with where you're heading, which works for a 2-hour flight but not a 33-hour train. Day 1 should say you're on the train, and New Delhi should only start the day you arrive. We're fixing it so long journeys show as "in transit" until arrival.
A gate change being flagged for me, with a fresh time to leave, is the moment this earns its place, since the last trip hinged on a gate change I only saw at the airport. Does Mojo need my booking emails, or can I add a flight by hand if I booked somewhere it never sees?
@karimbenkeroum Good question! So right now we have kept all the options open i.e.
Gmail sync, Forward an email, Upload booking, Add it manually, or tell Mojo (our AI agent) to add your flight.
It works not just for flights but for any bookings and then our agent is smart enough to properly cataegorize it and automatically start tracking it, if it is a flight
Congrats on the launch! How do you guard against Mojo hallucinating details like timings or entry requirements? Is there a way for the group to see where an answer came from before it goes into the plan?
@devansh_verma8
The AI never gets the last word on anything that matters.
- Facts come from sources. Flights, places and prices come from live data, not the model.
- Blank beats a guess. Missing fields stay empty, and code rejects anything that doesn't add up.
- Humans check. Our concierge team validates bookings before they're confirmed.
- You approve payments. Nothing is charged until you've seen the exact item and total.
- Easy to fix. Every error report becomes a permanent test.
If it's ever wrong, you'll see it, we'll catch it, and it won't cost you.
@devansh_verma8 @garg_rish quite thoughtful design pattern btw while you've explained where the data comes from, it's not always clear to the user which source backs a particular claim.
i noticed this on the Japan package i was exploring, where the visa card mentioned specifics like eVISA, five working days, and no printouts at the airport, but I couldn't tell what was sourced versus inferred.
Have you considered adding small source tags to individual facts, such as Live, Booking, Curated, or AI-inferred, supported with links where possible?
That would make verification much easier. For visa and entry requirements, authoritative sources like IATA Timatic could be particularly useful.