Tool-call filtering exposes the failed step only after the replay becomes hard to use.
Fast filtering protects the debugging flow, while complete token and cost data protects trust in the replay.
Who you’d be doing this for
“I know the bad tool call is in there, but the filter just keeps spinning while my incident channel waits.”
Darshan Arora · Backend Engineer
Uses conversation replay to diagnose failures in a production agent that calls several internal tools.
What is at stake
Filtered replay takes 8.4 seconds at P95 for large agent runs. You must weigh a faster query path against keeping token and cost breakdowns correct.
Why it isn’t already fixed
Every obvious fix costs something else. That’s the part you’d have to decide.
- replay speed vs. cost accuracy
- smaller payloads vs. debugging detail
- loading clarity vs. interface complexity
- routine rollout vs. high-volume edge cases
Why Sentry
Agent Tracing depends on developers reaching a failed tool call inside the conversation view.
Written with these in mind
Not your kind of problem? 20 more at Sentry, or browse every organization.
This is the setup. The work is inside.
Running it puts you in the room: the full situation and its constraints, stakeholders who push back in their own words, and the decisions that are yours to make. What you produce becomes a Day One Plan — work you can show someone instead of describing.