[All updates](https://qlane.ai/changelog)

Week 43 · Oct 19 – 25, 2026

# Everything that needs you, in one Inbox

Oct 25, 2026

The Inbox now holds what qlane can't test and the questions it asks, besides bugs. Pull-request plans check who may read a record, and a test account can hold a sign-in code.

The Inbox now holds everything in a project that needs a person's eye, sorted by kind. Bugs come with their history, area, evidence and every sighting. Can't test says what is stopping qlane from testing and who can fix it. Unconfirmed holds the bugs qlane could not confirm, and Questions holds what it asked you in chat or asked you to fill in.

Filter by kind, or by For you. Confirm a bug without filing it, or dismiss it with a reason and undo that later. In projects created since October 5, you can also move a pull-request bug to the Inbox with a note, linked to an existing issue if you have one. New entries carry a dot until you open them, and on a phone each entry opens on its own page.

* Pull-request test plans ask whether a change loosens who may read or change a record, and add a test that signs in as one account and opens another account's record. With only one test account stored, the pull-request comment says the access change went untested and offers Add a test account.
* A test account can hold a fixed sign-in code, for apps that ask for a one-time code after the password; qlane enters it when asked and hides it in logs and recordings. When the QA baseline can't sign in, it says which step failed.
* Your test values reach pull-request runs and every QA baseline flow, not only sign-up, and other form values are written in the page's language. The agent emails only qlane test mailboxes and never signs in to another service with credentials you didn't give it.
* A QA baseline flow can sign in as two to four of your test accounts in turn, for work like create, approve and execute. Flows your deployment switches off are listed apart, and a test never builds on a record another test made.
* A new Disposable app data setting on a test target lets QA baseline tests use records already in your app when it gives no way to create one.
* Fewer false bugs: a bug that cites a request or console line the agent never saw becomes an unconfirmed bug, and in pull-request runs a feature your app shows switched off is reported as blocked. The Test cases page marks a likely copy as Possibly a repeat.
* qlane follows your QLANE.md on how to do a step, such as which sign-up link or role to use, in QA baselines and in pull-request runs whose base is your default branch.
* Settings: testing level is a Faster–Smarter slider, rows fit on one line with a (?) for detail, and Archive has its own section.
* Get Started: with no source control connected, an admin connects it from the repository step, and a member is told which admin to ask. Sections no longer lock during setup once they hold cases or runs.
* The page where you enter the credentials or variables your coding assistant asked for over qlane MCP was reworked, and every qlane MCP tool has a readable title in your editor.
* The browser log kept with a session masks the codes and passwords a run made up, and everything the agent typed.
* Fixes: screenshots show in pull-request bug comments again, a repeated or late GitHub delivery no longer cancels the wrong run, attaching a .md, .csv or .json file no longer breaks a chat, a failed Jira connection attempt no longer breaks later filing, and you're no longer charged for a browser that failed to start.

[Older · Week 42 → One Inbox for every bug](https://qlane.ai/changelog/one-inbox-for-every-bug)