List runs via MCP, typed tool schemas, mobile test cases, and API key labels

Highlight
Featured this release
PR review bot keeps one comment per pull request
The review bot now edits a single sticky comment in place rather than posting a new comment on every round.
Previously, each review round – queued, running, result – added a fresh comment to your pull request. The bot now maintains one comment per PR and updates it as the run progresses, with a collapsible history of earlier rounds at the bottom. The GitHub review verdict (approve, request changes, or comment) is still posted as a separate review object and links back to the sticky comment.
# Direct link
Highlight
Featured this release
Test Cases page works on mobile
The scenario tree and test list are now usable on phones – no horizontal scrolling and menus are always reachable on touch.
On narrow screens, the scenario tree slides in as a sheet triggered by a selector button instead of squeezing into a thin sidebar. The table fills the full width, drops columns that don't fit, and moves last-run history directly under the test name so it stays visible. The row options menu is always visible on mobile – previously it was hidden behind hover and never appeared on touch devices.
# Direct link
Update 03 of 11
Agents can list runs via MCP and API
A new list_runs MCP tool and GET /v1/run endpoint let agents query runs by app, plan, status, branch, or date range – no human lookup required.
list_runs is now available in the QA.tech MCP, letting agents query runs by application, test plan, status, trigger, branch, commit hash, and time window. It returns shortIds that feed directly into get_run and rerun_run, which previously required a human to paste a run ID. A matching GET /v1/run endpoint is also available in the public API.
# Direct link
Update 04 of 11
Chat assistant explains when file downloads aren't possible
The assistant tells you clearly when a request isn't supported, instead of attempting it and running into limits.
When asked to produce a CSV export or other file download, the chat assistant previously tried to work around the limitation – sometimes producing chunked output or stopping mid-way through a large result. It now tells you upfront that file downloads and attachments aren't supported in the chat surface, and offers inline tables or short summaries as an alternative where they fit.
# Direct link
Update 05 of 11
MCP tools now expose typed schemas for all parameters
start_run, append_chat_message, and the mobile build upload flow are now self-describing – agents can build correct payloads from the schemas alone.
start_run and append_chat_message previously exposed applications, notifications, and applicationOverrides as opaque arrays. They now carry full typed schemas matching the tool contract. The two-step mobile build upload (application_build_upload_url → HTTP PUT → create_application_build) is documented in both tool descriptions with cross-references, and the upload-URL response includes a nextStep hint pointing to the next call.
# Direct link
Update 06 of 11
Name your API keys
You can now give each API key a label so it's easy to tell them apart.
API keys in Settings now have an optional name field. Set a name when you create a key, or edit any existing key to label it – handy when you're managing separate keys for different environments or integrations.
# Direct link
Update 07 of 11
Network settings shows the correct outbound IPs URL
Settings → Network now links to https://api.qa.tech/v1/outbound-ips instead of the old path.
Settings → Network now links to https://api.qa.tech/v1/outbound-ips instead of the old path.
# Direct link
Update 08 of 11
Personal API keys
You can now create and manage your own personal API keys from the dashboard.
The Settings → API keys page now lets individual team members generate personal API keys in addition to any project-level keys. Personal keys are scoped to your account, shown once at creation time, and can be revoked from the same page.
# Direct link
Update 09 of 11
Tighter PR review status comment handling
Each pull request now gets exactly one status comment, updated in place as runs complete.
The status comment writer has been tightened so each PR gets exactly one status comment, updated in place as runs complete.
# Direct link
Update 10 of 11
Sign-up page is faster
An analytics timing issue that slowed the sign-up page has been resolved.
Analytics tracking on the sign-up page was running synchronously and adding noticeable latency on load. Tracking now runs asynchronously, so the page renders without delay.
# Direct link
Update 11 of 11
Simpler sign-up when arriving from a shared link
Visitors following a shared run or project link now see a focused sign-up form instead of the full marketing layout.
When an unauthenticated visitor opens a shared link (a /p/… URL) and reaches the sign-up page, the page now shows a streamlined layout with just the form. Testimonials and marketing content are omitted. The full layout is unchanged for direct visits.
# Direct link

PR review bot keeps one comment per pull request

Test Cases page works on mobile

Agents can list runs via MCP and API

Chat assistant explains when file downloads aren't possible

MCP tools now expose typed schemas for all parameters

Name your API keys

Network settings shows the correct outbound IPs URL

Personal API keys

Tighter PR review status comment handling

Sign-up page is faster

Simpler sign-up when arriving from a shared link

Your team moves fast. Can your testing keep up?