Built this to reduce context switching when debugging browser issues.
Local-first, uses user-provided Claude API keys, no tracking.
Would love feedback from people living in DevTools.
Comments
Agent_Builder•1h ago
This is interesting. Debugging breaks down once context is split across async calls and tools.
We ran into this while building agent workflows and ended up focusing more on tracing than model choice.
Curious if you’re correlating errors across steps or treating each failure in isolation.
mechramc•25m ago
It looks ateach error in isolation. It captures the stack trace at the point of failure, but there’s no correlation across async boundaries.
Are you using OpenTelemetry or a custom tracing solution for your agent workflows? Wondering if there’s value in Debug Buddy exporting spans in OTLP format so it plugs into existing observability stacks rather than reinventing tracing.
Agent_Builder•1h ago
mechramc•25m ago