Step-level traces
A waterfall view of every step with timing, so you can spot the slow tool call or runaway loop instantly.
Every run is traced end to end — prompts, tool calls, tokens, latency, and cost. When something goes wrong, replay the run deterministically and find the cause in seconds, not hours.
A waterfall view of every step with timing, so you can spot the slow tool call or runaway loop instantly.
Full inputs and outputs for each step, searchable and filterable by agent, status, model, or custom tags.
Per-run and per-agent spend broken down by model. Set budgets and get alerted before you blow past them.
Re-run any production trace with the same inputs and tool responses to reproduce a bug exactly.
Trigger on error rates, latency spikes, or cost thresholds — routed to Slack, PagerDuty, or webhooks.
Attach scores and human feedback to runs, then build datasets to evaluate prompt and model changes.
AgentArc traces are OpenTelemetry-native, so they flow into the tools your team already watches.
Get full traces on every run from your very first deploy — free.