Developer Deploys Claude Agent to Monitor and Auto-Fix Project Errors

and ofc this is a milestone on the way to a autonomous product, but we aren't there yet

@ilakkManoharan I already have it - it's custom per project

@GregKamradt does claude read your logs every X minutes and then goes to sleep? or do you have enabled hooks for this?

@GregKamradt I would be delighted to build this agent for you.

@GregKamradt The step I'm on in this flow is merge gating. Ideally the end result should be that the user intents are met without prompting. And like you said, you shouldn't be in the way. There should be minimal activation energy between your desire and the agent output.

@GregKamradt mmm, can you point me to your api endpoints, for no particular reason at all 😅👀🥰🤫

@codegraph yep agree

@notismaelvega It's a cron, and no need for every X minutes

the side project only cron's 2x time a day, so claude does the same

I don't want to overload it

@GregKamradt My agents are custom too…

@GregKamradt I joke ofcourse ... but I hope those instances are sandboxed!

@GregKamradt this is it. agent handles noise, you handle decisions. you just bought back enough time to actually ship.

@GregKamradt This middle ground is underserved. Full autonomy gets the hype, but supervised delegation with structured reporting is what actually works in practice. How are you benchmarking fix quality over time?