v2026.09.10
TL;DR
Answer Thunders AI or connect a tool without leaving the chat, a new Read-only role controls project access, and admins get new controls for feature flags and catching an outdated browser engine.
FEATURE
- Never leave the chat: Thunders AI's clarifying questions now show up in a dedicated dock above the chat box, with suggested answers.
- Connect and enable tools right from the chat: Thunders AI can offer to turn on Web Browse or connect a tool like Jira, Xray or Linear, right from the chat.
- Variables that update mid-run: a step can now change a project or environment variable's value, or clear it, while a run is in progress. The new value applies globally, so every other test case or run that references the variable sees it, not just the one that set it, and the run view shows how stale the current value is.
- Read-only role: Admins can now give a person the Read-only role for a project. This role lets someone view a project without changing it. Admins manage this access from a new per-project access table.
- RGAA accessibility audit grid: Users can now download the official RGAA 4.1.2 audit grid in ODS format. This sits next to the existing PDF accessibility report.
- Feature flags from configuration: admins can now set feature flags directly from their own configuration.
- Thunders MCP: the Thunders MCP surface now offers a delete_step tool, so an MCP client can remove a test step directly.
- Xray auto-sync: turn on the new auto-sync toggle for a project and mapped test cases push their pass or fail results to Xray automatically as soon as a run finishes. The manual sync button still works for a one-off re-sync.
IMPROVEMENT
- MCP Execution Mode: test runs now show MCP as the execution mode for a run created through Thunders MCP.
- Thunders MCP: an action taken through Thunders MCP now shows up in a test case's activity feed, naming the person who did it when they connected with their own account.
- Test Assets list: the list now loads with server-side paging, so large lists load faster. The page-size dropdown is back too.
- Extract File step: the step can now skip empty rows and cap the number of extracted rows.
- Accessibility scans: scans now resolve an element's accessible name from every source the standard defines, not just its alt text. Admins can also opt additional experimental accessibility rules in by id.
- Extract API response step: a step can now save a value from an API response directly into a project or environment variable.
FIX
- Visual Compare accuracy: Visual Compare's diff crops and overlay boxes are now more accurate.
- Thunders AI chat: fixed a duplicate assistant message appearing twice in the same reply.
- Sequential runs: a variable a prerequisite test case creates is now shared with the rest of the sequential run.
- Browserless isolation: a self-hosted install no longer falls back to Thunders' own Browserless fleet when its configured instance is unavailable.
- Thunders AI reliability: fixed an issue where Thunders AI could fail partway through a task, after using more than one tool in a step, when your configured AI provider was Gemini.
- UI polish: removed the engine name from accessibility PDF report headers.
Ready to ship faster with smarter testing?
