Thunders Copilot is now Thunders AI, custom MCP connectors, and soft-delete test runs
TL;DR
Thunders Copilot is now Thunders AI. This release also adds custom MCP connectors with a redesigned connectors page, gives Thunders AI a feedback loop plus the ability to file bug tickets for you, and lets you soft-delete test runs.
FEATURE
- Custom MCP connectors: connect your own MCP servers to Thunders AI, alongside the built in integrations, from a redesigned connectors management page.
- Report bugs using Thunders AI: ask Thunders AI to file a bug ticket for failed steps mid conversation. It creates the ticket in your connected provider and links it back to the run.
- Thunders AI feedback: rate a Thunders AI response with a thumbs up or down right in the conversation.
- Soft-delete test runs: delete test runs without losing them for good. Deleted runs are hidden from lists and stay recoverable.
- Apps get a full details page: Apps now open in a full page like Environments
IMPROVEMENT
- Thunders Copilot is now Thunders AI: building the experience into a colleague you can tag anywhere to get your testing work done.
- Dropdown elements handling: the new dropdown interaction handling is now enabled by default on all test runs.
- Personal integration credentials: connect to integrations at the personal-project level instead of being shared across the whole organization.
FIX
- PDF report exports: test run PDF reports now export the full document instead of just the first page, keep step numbering aligned, use the right font, and render org logos instead of failing.
- "Generate random" steps: steps that generate a random value now actually produce a new value each run, instead of repeating the same one.
- Natural language scroll: a step written as "scroll up" or "scroll down" is now recognized as a scroll action instead of unknown.
- Interrupted vs. failed runs: a run whose browser process crashes is now marked interrupted instead of failed, so it is easier to tell a real bug from an environment hiccup.
- Thunders AI fixes: stopped Thunders AI refusing to make edits after an in-editor test run, and fixed it failing to browse websites during a conversation.
- UI polish: fixed a flickering report popover, blank list grids on some pages, global hotkeys double-firing over open dialogs, a Firefox crash on logout, retry groups not loading their attempts, invalid app names being accepted, and billing copy that did not match the website.
Ready to ship faster with smarter testing?
