Skip to content
TestAutomate Join the waitlistWaitlist

Get notified at launch

TestAutomate isn't released yet. Leave your name and email and we'll notify you when it launches.

Home / Changelog CHANGELOG

What shipped, dated and honest

No vaporware, no backdated launches. If it's listed here, it's in the product. If it's not shipped yet, it's marked "coming" — here and everywhere else on this site.

— Per-workspace test storage, per-tenant run locks

Tests moved out of shared storage into fully per-workspace storage, and every run now takes a per-tenant lock. Cleaner isolation between orgs, and a busy neighbor can never contend with your sweep.

— CI: boot smoke test on every commit

Every commit now boots the full stack — Node, Postgres, migrations — in CI and must serve before it merges. Boring and load-bearing: a QA company whose own main branch doesn't boot has no business grading yours.

— Provider registry: Gemini and Groq tiers

The model provider registry now includes Gemini and Groq alongside Anthropic and Bedrock, with per-tier routing. Cheap models read pages, stronger models judge failures — and you bring whichever key you already have.

— Voice input rebuilt

Dictating a test is now a first-class authoring path, not a gimmick. Rebuilt capture and transcription flow, so "say the test out loud, watch it become YAML" works the way it should.

— Hard budget ceilings and live spend meters

Every workspace gets a hard model-spend ceiling and a live meter on every run and sweep. When the ceiling is hit, work pauses cleanly and resumes when you raise it — no surprise invoice, ever.

— Verification sweeps: durable queue and warm resume

Sprint UAT sweeps now run on a durable queue. Kill the process mid-sweep, restart, and it resumes from the last completed ticket instead of starting over. Long boards survive real-world interruptions.

— Escalate-on-fail ships

A failing run is now automatically re-driven by a stronger model before it is reported. False alarms drop, and every red verdict you see has been checked twice by construction.

— Blocked is not failed, and flake labeling

Verdicts split into pass, fail, and blocked: broken staging or an unreachable environment never counts as a product bug. Tests that pass on retry are labeled as flaky instead of quietly going green.

SHIPPING NEXT

On the bench, not yet in the product

These are being built now. Nothing below is claimable today — that's the point of listing them separately.

1

Scheduled runs + report delivery

Nightly suites and sweep reports delivered to email or Slack, so verdicts find you instead of the other way around.

coming
2

Cloud browsers + parallel runs

Managed browsers with live view and takeover, running suites in parallel — alongside, not instead of, your own Chrome.

coming
3

CI trigger + GitHub Action

Kick off a targeted sweep from CI and check a PR's blast radius before merge.

coming

Watch it ship, or help steer it

Design partners see every release first — and decide what comes next.