I built a way to see your app, AI calls and errors in one trace

Solo founder here. When an AI feature breaks in prod, the error is in Sentry, the spans are somewhere else, and the model call is in a third tool. axonpush puts them together: point your existing OpenTelemetry exporter or Sentry DSN at it, no new SDK app spans, exceptions and model/tool calls on one trace; errors grouped into issues hand a failing run to Claude Code or Codex with the trace attached optional gateway for AI traffic: block or redact risky tool calls and PII, and cap spend per agent (falls back to a cheaper model instead of killing the run) It's early, and I'm looking for the first few teams to try it. I'll set it up with you personally. What would make you trust a new tool with your production traces? That's what I'm trying to figure out. axonpush.xyz