Use cases
Different questions.
One recorded path.
Choose the job you need to get done. Each journey maps to a capability that exists today, not a promised feature or a fictional customer result. Some depend on your plan and capture settings.
01 / Agent developers
Find the step behind a failed outcome.
The request returned, but you still don't know which model or tool action made the workflow go wrong.
What the product makes possible
- Open captured spans as a waterfall, transcript or agent lanes.
- Inspect a recorded failure and, where evidence allows, compare it with an earlier run that has no recorded failure signal.
- On Builder and above, export a regression fixture when input capture is on and your role can manage datasets.
02 / Platform and reliability teams
Make agent traffic observable and operational.
Failures and retries need to be understood across environments, releases, models and services.
What the product makes possible
- See recorded costs, errors, latency and tool calls.
- Inspect gateway guardrail decisions, warnings and overhead.
- Use the available SLO, trace and incident views to narrow investigations.
03 / Risk and audit teams
Verify a specific recorded range independently.
A dashboard is not sufficient evidence that a particular gateway event has not been silently altered.
What the product makes possible
- Inspect a hash-chained, signed audit range.
- Use the documented offline verifier and inspect trust boundaries.
- Read the published limitations and distinguish gateway events from OTLP-only spans.
Working on a real agent rollout?