Cursor Agents
AI agents that write code AND prove it works
Start new thread
Raghav Mehra

Cursor Agents - AI agents that write code AND prove it works

by•
Autonomous AI engineers that spin up cloud VMs to write, test, and visually verify code. Assign a task. Get back a PR with a screen recording of the agent testing it. Finally, AI that closes the loop instead of handing it back.

Add a comment

Replies

Best
Raghav Mehra
Hunter
📌
I've been watching the agentic coding space closely, and the pattern I keep seeing is this: the AI writes the code, and then a human has to pick up the last mile. Run the build. Check the UI. Verify it didn't break something upstream. That part never got automated. It just got disguised. What made me stop and look twice at Cursor Cloud Agents is the screen recording. The agent literally records itself navigating the UI and attaches it to the PR. That's not a feature, that's a different relationship with AI-generated code. The parallelism angle is also underrated. Being able to assign multiple tasks to separate cloud VMs simultaneously, without competing for local resources, is the thing that finally makes "managing AI like a team" feel real rather than theoretical. I'm genuinely curious whether the community sees the verification behavior as the core unlock, or whether the async delegation from mobile is what moves the needle for founders specifically. Keen to hear from the Cursor team too, what use cases have surprised you most in early testing?