Itineraries that survive contact with a real city
One district per day, sequenced by distance. Opening hours enforced instead of guessed. Budgets split into stay, food and admission — with no invented flight cost.
London itinerary, planned district by district
Get a London itinerary where every stop is a real place with a real opening time, the days do not zig-zag across town, and the budget is broken down instead of invented.
4 size presets →Paris itinerary, planned district by district
Get a Paris itinerary where every stop is a real place with a real opening time, the days do not zig-zag across town, and the budget is broken down instead of invented.
4 size presets →Tokyo itinerary, planned district by district
Get a Tokyo itinerary where every stop is a real place with a real opening time, the days do not zig-zag across town, and the budget is broken down instead of invented.
4 size presets →Itinavia builds city itineraries the boring way: work out where the places are, put each day in one district, sequence the day by distance, check the opening hours, and only then write the prose.
Most AI travel planners get this backwards. The model is asked to choose the places, order them, time them, price them and describe them in one pass — and it can only do the last one well. The result reads beautifully and falls apart on the ground: a day that crosses the city twice, a museum that is shut by the time you arrive, and a budget with a flight cost in it even though you never said where you were flying from.
Here the first four decisions are made by a deterministic planner reading OpenStreetMap. It picks places that have a Wikidata entry, ranks them by how widely they are documented across Wikipedia languages, clusters them geographically so each day gets one district, and sequences that district so you are not doubling back. Opening hours are enforced where the data has them. The budget is split deterministically into stay, food, admission and local transport — and there is no flight line, because without a departure city there is no honest number to give you.
Then a language model is allowed to write the wording, and only that. Its output is checked against the facts it was given: if it adds a price, a time, an opening hour or a link that was not in the plan, the whole text is thrown away and the deterministic version is used instead.
Start with a city: Tokyo, Paris, London. Each has plans for two and three days, for solo travellers and for pairs. If you want to know exactly how the plans are built, read how it works.
Frequently asked questions
How is this different from asking a chatbot for an itinerary?
A chatbot writes fluent prose about places it cannot check. Here the places, their order, the timings and the budget are decided by a deterministic planner reading OpenStreetMap, and then a separate quality gate verifies the result: no day crosses town twice, no district is repeated across days, no day runs past a working day's length, and no price appears that was not in the data. Language models are only used to polish the wording, and any wording that contradicts the data is thrown away.
Why is there no flight cost in the budget?
Because you did not tell us where you are flying from, and a made-up number in a budget you plan against is worse than a missing one. The budget covers the ground: accommodation, food, admission and local transport.
How do you decide which places to visit?
Places are pulled from OpenStreetMap with a Wikidata entry, ranked by how widely they are documented across Wikipedia languages, then clustered geographically so each day stays in one district. Museums and temples that close at 17:00 are never scheduled into an evening slot.
Can I trust the opening hours?
When a place carries an opening-hours tag we enforce it: a place shut for the whole week is dropped from the pool, and nothing is scheduled into a closed period. When the tag is missing we fall back to the typical hours for that type of place — so treat the times as a planning guide and confirm anything time-critical with the venue.