You can also browse, search and filter all release notes directly in the application → Open release notes in TAI.
Admins get a clearer label for the team's personal agents
Improvement. Only Company Admins.
On the Agents list, the personal-agents section was labelled "My agents" for everyone. As an admin you actually see every user's personal agents there, not just your own, so that label was misleading. It now reads User Agents for admins, with a short note explaining that a regular user only ever sees their own here. Everyone else keeps the familiar "My agents".
Agents compute dates and totals with exact built-in calculators
Improvement.
Date and number work in agent answers — nights between stays, gaps and overlaps in an itinerary, price totals against a stated sum — is now routed through built-in deterministic calculators instead of being worked out by the language model itself. That removes the main source of off-by-one night counts and slightly-off totals.
Documentation pages have a cleaner, easier-to-read layout
Improvement.
Documentation articles now sit on a dedicated reading surface — a bordered card centered on the page, with roomier spacing and a clearer heading rhythm — instead of running flat against the background. It's easier on the eyes in both light and dark mode.
Itinerary Checker: clearer results without unreliable scores
Improvement.
The Itinerary Checker no longer shows a 0-100 health score, a single overall PASS/WARN/FAIL verdict, or a per-passenger completeness percentage. Those numbers were worked out by the model itself and weren't reliable — the same booking could produce different figures from one run to the next — so they added false precision without adding trust. The report now leads with what matters: each finding with its own severity, a per-category verdict, and, for passengers, exactly which required fields are missing — so you act on the specifics instead of headline numbers that didn't hold up.
Report Back now returns you to the list you came from
Fix.
Opening a report and pressing Back used to reset you to the default All reports view, losing your search, filters, grouping, and sort. Back now returns to the exact list you left — including when you open a run from a schedule's own page, which now returns you to that schedule.
Long report sessions no longer error when exporting a PDF or HTML report
Fix.
In a long agent session — for example a large report refined over many steps — asking the agent to generate a PDF or HTML report could fail with a validation error. When trimming the conversation to fit the context window, the platform no longer corrupts the parts of the history it keeps, so these exports now succeed.