cursor background agents are truly next level

Thank you ! I was literally saying this to @samuellhuber.eth today

+ the slack integration is underrated

Farcaster Development Boutique @ https://dTech.vision 💜 in love with tech 🤓decentralization, fitness and oldschool rap 😎

Full-stack digital product consultant @ BoundCorp.net

Be steadfast. Amor Fati.
Building @tipdotmd

@leewardbound how do you use them?

From what I gather I'd now write a detailled spec. 

Then just let agents run run run

instead of writing a spec and having claude code or so watch me

How do you verify? Backend I can have tests that verify itself UI is harder imo

they handle all small and medium maintenance tasks for me, not so much UI dev at the moment (tho i believe they do open a port so you can browse the UI?)

i can ask for a refactor or a cleanup or a feature tweak (or all 3 at once in 3 different agents), check in once or twice over the next hour with suggestions, and then PR is autocreated for me. the big unlock for me here is not having to make PRs and manage branches, i can get quick little bug fixes in without context switching.

@sayo shared this that can work for UI test that is MCP based but not sure if can run in a background agent.

Personally been testing myself the UI after work is done, list all issues I find in a plan then send again to fix. 

Managed to build pretty cool stuff this way

Yes CLI fallback is usually what I instruct agents to do, for ex using Taskmaster.

background agents currently unable to use MCP at all afaik but you can usually patch together CLI instructions to get much of the functionality back

@sayo shared this that can work for UI test that is MCP based but not sure if can run in a background agent.

Personally been testing myself the UI after work is done, list all issues I find in a plan then send again to fix. 

Managed to build pretty cool stuff this way

https://youtu.be/QaL3fY5q2iQ?si=eaHetza00sdsjzDQ