Observability for Agent Runs

About this workflow

Instrument your agent like production software: traces, session replay, cost tracking, and failure analysis on open standards.

What happens when you run it

1. 1
   The trace tree is the run's skeleton - build it first.

   Runs flow-observability-agent-runs-1-instrument-spans-across-the-loop on Claude Sonnet 4.6
2. 2
   The debugging superpower: watch the run again, slowly.

   Runs flow-observability-agent-runs-2-build-session-replay on Claude Sonnet 4.6
3. 3
   Cost per session shapes design decisions - surface it everywhere.

   Runs flow-observability-agent-runs-3-track-the-economics on Claude Sonnet 4.6
4. 4
   Observability proves itself on the next bad day - rehearse it.

   Runs flow-observability-agent-runs-4-run-the-failure-analysis-drill on Claude Sonnet 4.6

Best for

• •Recurring tasks you currently copy-paste between tools.
• •When a single prompt isn't enough — you need a chain where output(n) feeds prompt(n+1).
• •Fork and adapt if your inputs differ materially from the author's use case.

Models used