Loading AgentsKit ecosystem navigation…
AgentsKit / execution lab
SIMULATION · NO LLMSource ↗

CONF42 OBSERVABILITY 2026

Don’t trust the answer.
Follow the evidence.

One incident. Two execution paths.
Every decision, failure and recovery — visible.

Checkout API
is failing.

“Errors increased after the latest deploy. Investigate and propose the safest next step.”

5xx error rate18.4%
Controlled fixture databaseline 0.2%
Service
checkout-api
Deployment
v2.8.4 → v2.8.3
Instructions
runbook-v3
Context
snapshot-v7

Ready to investigate

Start the investigation. The trace will reveal what a final answer cannot.

Awaiting execution0 events

An answer is not a trace.

Run the incident to see context, tool boundaries, policy and recovery unfold.

Application trace · actual fixture execution