Every API release from the v1 days to the current v3.4.0, newest first. Breaking changes are marked — we give 90 days' notice and usually a migration script.
v3.4.0
/predictions/delay now returns confidence intervals, not just a point estimate. The point estimate is unchanged — old integrations keep working.
Aircraft rotations added to /routes responses (which tails fly which legs, per week).
GraphQL: pagination cursors on all board queries.
v3.3.0
Baggage carousel and belt assignments for 61 airports that publish them machine-readably. We'd love to say all airports, but most just... don't. The field is null where unavailable rather than absent, so schema consumers stop guessing.
Webhook replay: Business plans can replay up to 24 hours of events by topic.
v3.2.1
Fixed a caching bug where a diverted flight's new arrival airport could briefly show the original one. One customer found it in an hour; we bought them coffee.
v3.2.0
Schedule coverage extended to 12 months forward (previously 6) across all 3,900 airports.
Carrier change files exposed via /schedules/{carrier}/changes for nightly delta pulls.
Position data freshness improved: p95 ingestion lag down from 9s to 4s after a receiver-network expansion.
v3.1.0
Swift SDK 2.0 with async/await throughout.
Rate limits raised across all plans at no charge — Hobby 30→60 req/min, Pro 300→600. Capacity grew; sharing felt right.
Deprecated the X-Api-Key header in favor of Authorization: Bearer (the header kept working until v3.4's cleanup, and yes, a few of you got a friendly support email before it went).
v3.0.0
Breaking: v3 base URL (/v3), Bearer auth as the primary scheme, and a unified flight object across status, schedules and routes. The v2 API ran in parallel until 2025-08-31 — every v2 customer got a migration guide, a shim library and, where asked, a video call. About 40 integrations moved in the first week and the rest trickled in. Two never migrated and let their plans lapse; we think of them sometimes.
Delay predictions generally available (the model had run in shadow mode on v2 since November 2024 — five years of ASQP data, weather, connect-through load).
Performance patch: search queries on large airports (ATL, LHR, PEK) were 2–3× slower than spec after a November index change. Found by a customer's monitor, again. Buy stock in customers with monitors.
v2.4.0
Route networks endpoint: flown routes, frequencies, load estimates per carrier.
Airport delay index (rolling 24h) on /airports/{icao}.
v2.3.0
Webhooks moved from beta to GA: signed payloads, 12-hour retry, 10 endpoints per account.
Go SDK 1.0.
v2.2.0
Schedules endpoint (6 months forward) and carrier metadata.
Sandbox environment with recorded fixture data — free on every plan.
v2.1.0
Status search: query flights by airport, route or status instead of one id at a time.
Rate-limit headers on every response (they'd been on some; now all).
v2.0.0
Breaking: v1 (a hobby-grade JSON dump with delusions of grandeur) shut off on 2024-09-30. Migration was mostly a base-URL change plus renaming flightno to id. Everyone made it.
Fresh architecture, one flight object, real auth. The v1 API was three PHP scripts and hope; we don't miss it.
Live tracking via ADS-B receiver network plus carrier integrations.
Pre-2024 v1.x releases live in the Git history and, honestly, nowhere else — nobody has ever asked. Current health: status page. Release notifications go out via the dashboard and a very low-traffic mailing list.