Trace Autopsy
Watch exactly where the agent went wrong. Every step, exposed.
agent traces are the single most opaque artifact in AI engineering, seen only by practitioners; this opens the black box of 'what did the agent actually do' to a total beginner, for free, in one click
What goes in, what comes out
Learner steps through a real, anonymized, seeded agent execution trace one tool-call at a time — plan, tool call, result, next plan — flagging 'healthy' or 'this is where it broke' at each step. The tool reveals the actual root-cause step afterward and contrasts it with the visible symptom step most learners mis-flag.
click through trace steps, flag the one step believed to have caused the failure
root-cause reveal vs the flagged step, a symptom-vs-cause explainer, an aggregate 'most learners flagged step N, the real cause was step M' stat
trace-library picker by failure type (loop, wrong-tool, hallucinated-step, scope-violation), step-by-step vs full-trace-then-diagnose mode
FREE, top-of-funnel — /playgrounds gallery; the identical trace-reading skill is graded inside Agent Builder and MCP Playground L2 labs
Concepts click when you open the machinery.
Three labs are already live and free — the same hands-on style this playground brings to its module.