1 · Capture
One run, one context
Cost, model calls, retries and limit decisions are tied to the same run.
Projects / KazenAI Run receipts
Run receipts are part of KazenAI Agent FinOps. They connect each cost and limit decision to the calls, retries and incidents behind it, so you can explain an expensive or failed run.
You work with the person who built it. The first 30 minute call is free, and if I can’t help, I say so before you pay anything.
Decisions without a receipt0
Automated test on PostgreSQL, not customer traffic.
You may already have provider logs, application logs and traces, and still spend hours joining them during an incident.
“What actually happened in this run?”
A clear path through the captured model calls, retries and limit decisions for a given run.
“Is this the same failure as last week?”
Recurring evidence can be grouped, assigned and reviewed, with the reason recorded.
“Did the fix change anything?”
Compare two captured runs under stated conditions. A comparison can show a difference; it doesn’t prove the cause.
“What if the cost data is missing?”
Missing or stale cost evidence is not silently changed to zero.
Explain a costly run without stitching logs together.
Run timelines and incident records use the same run context as Agent FinOps.
1 · Capture
Cost, model calls, retries and limit decisions are tied to the same run.
2 · Investigate
Move from an alert to the costly step, retry or limit decision behind it.
3 · Record
Ownership, status and reason are kept, so your team can review the decision later.
Illustrative incident evidence from a captured run.
workflow:report_generation run_id:run_01J... incident:repeated_model_retry occurrences:4cost_state:settledreview_status:needs_review comparison:existing_runs_only
Incident workflows are verified locally with synthetic data. Customer production use requires separate verification.
Comparing existing runs is evidence, not causal proof or an automatic fix. Hosted evidence uses sample data.
The first 30 minute call is free, and I reply within one business day.