Where did the run spend time?
Use the execution timeline to identify slow model calls, tool calls, retries, and waiting periods.
Trace analysis for AI systems
Make OpenTelemetry and OpenInference traces useful for debugging.
TraceProbe gives developers a focused desktop view of AI agent execution, from high-level reasoning flow to individual tool calls and timing details.
See the shape of an agent run.
Trace data is valuable when it explains how a run unfolded. Use the reasoning graph to follow decisions, tool activity, branches, and handoffs without losing the context of the original execution.
- Review reasoning and tool relationships visually.
- Connect trace spans to the step that produced a failure or delay.
- Keep sensitive logs and imported run data local on your device.
From trace events to engineering decisions
Use the viewer to answer the questions that matter during an investigation.
Which step changed the outcome?
Follow the reasoning graph and surrounding events to isolate the branch or tool result that redirected the run.
What should we test next?
Turn the observed failure mode into a focused prompt, tool, model, or orchestration experiment.
Built for local investigation.
TraceProbe is designed for developers who need to inspect agent behavior without copying sensitive trace content into a hosted dashboard. Open sample runs, import supported logs, and keep the analysis close to the machine where you work.
Inspect your next AI agent trace.
Install TraceProbe and start with the built-in sample runs.