What production context do your coding agents actually need?
We’re launching Appxiom on Product Hunt today, and I wanted a place for a question that kept coming up while we were building it.
Coding agents can read your whole repo, but they can’t see what breaks in production. So when a bug lands, someone pastes in a stack trace, a log snippet or a user complaint, and the agent works from that.
We’ve been building the other side of it: crashes, freezes, memory leaks, failed network calls and broken user journeys, handed to the agent over MCP. Along the way we kept disagreeing about how much the agent should see. So I’d like to hear how you handle it:
When you ask your agent to fix a bug today, what do you paste in by hand?
Is the technical error enough, or would you want the agent to know which user flow broke (signup, checkout) and how many users it cost?
Should the agent decide which bug to tackle first, or should a human always make that call?
If you’d like to see our approach, the launch is here: https://www.producthunt.com/products/appxiom?launch=appxiom
I’ll be reading every reply today.


Replies
Excited to try it out 🚀