Question
Short answer
Tracify records every prompt, model call, tool call, retry and error in an AI workflow as one trace, so you can open a bad answer and see which step caused it. Add the TypeScript or Python SDK, or send OpenTelemetry spans, and runs show up in the trace viewer.
In more detail
Each step keeps its input, output, timing, token counts and cost, next to the model or tool that produced it.
Start at the failed run, open the trace, and jump to the first step that went wrong instead of searching through separate logs.
Install the SDK, wrap your model client, and run your workflow once. The first trace appears within seconds.
