Post-run AI analysis, geo-routing, and self-describing MCP tools

Highlight
Featured this release
Post-run AI analysis and automatic reruns
After each run completes, an AI agent classifies failures, re-runs flaky tests, and shows a full timeline in a new Analysis tab.
When a run finishes with failures, QA.tech classifies each failing test as a real failure or a transient/platform flake and re-runs candidates automatically. Set the budget under Project Settings → General → "Automatic reruns" (Off, or 1–3 attempts per test per run). The "Analysis" tab on the run results page shows a chronological timeline of per-test classifications with reasoning, which tests were re-run, and settled outcomes. When the agent has a specific recommendation – like splitting a test that keeps hitting the step limit – it shows a "Fix in chat" button that opens the assistant prefilled with context.
# Direct link
Update 02 of 05
Bitrise CI/CD integration guide
New docs page for integrating QA.tech into a Bitrise pipeline for APK, IPA, and web app builds.
Covers Android APK and iOS IPA upload-and-run flows, a plain HTTP trigger for web apps, and a polling recipe for using QA.tech as a release gate before a store submission. Linked from the CI/CD integration overview on docs.qa.tech.
# Direct link
Update 03 of 05
Test from specific countries using geo-location routing
Device presets now accept a region – US, UK, or Germany – and route the test browser through a residential proxy in that country.
Select a region on any device preset to run the test through a residential/ISP proxy in that country. Locale and timezone follow the region automatically, so you get a full market simulation without configuring them separately.
# Direct link
Update 04 of 05
MCP tools now describe return shapes and next steps
Every create_* and start_* MCP tool now documents what it returns and which tool to call next.
Each tool's description ends with a Returns: { ... } line matching the actual response shape and, for async operations, a Next: hint pointing at the right polling tool – for example, start_run → get_run. AI coding agents can chain calls without a discovery round-trip.
# Direct link
Update 05 of 05
PR review bot posts new GitHub comments for status updates
QA.tech now posts a new comment on a pull request for each status update, instead of editing the previous one.
GitHub notifies reviewers on new comments but not on edits – status updates now show up in the notification feed.
# Direct link

Post-run AI analysis and automatic reruns

Bitrise CI/CD integration guide

Test from specific countries using geo-location routing

MCP tools now describe return shapes and next steps

PR review bot posts new GitHub comments for status updates

Your team moves fast. Can your testing keep up?