I’m really fed up with automatic verifications triggering; they burn through three times as many tokens and take forever to complete any simple task or change I request.
This has been happening since the last few updates.
In Agent window: agents trigger sub-agents with “computer use” capabilities that perform screen recordings for verification—even when I haven’t asked for them.
In IDE window: browser controls trigger to verify changes and take a huge amount of time.
Please, is anyone else experiencing this? How can I prevent this from happening without my consent?
Hey @adri9deu, nothing’s misconfigured. Two separate switches.
IDE browser checks: Cursor Settings → Browser → set Browser Automation to Off, then start a new chat. To keep it but only on request, leave it on and turn on Browser Protection so the agent asks first.
Agents window computer-use and recordings: those are Cloud Agents. Add /no-test to a prompt for one run, or set {"enable_testing": false} in the repo’s .cursor/environment.json (or cursor.com → Cloud Agents → Environments → your repo → Disable Testing) to turn it off for that repo.