How this works

AI assistant for Dhavan’s published portfolio material. Answers can be incomplete or wrong, so do not submit confidential code or data.

Your questions go through Hugging Face, and a model run by DeepInfra writes the answers. Chats may be retained for about 30 days to operate, debug and improve the assistant; DynamoDB TTL deletion is approximate. Cal.com handles call bookings. Details are in the privacy policy.

Use the suggested questions or ask about published projects, credentials, and engagements.

Projects / KazenAI Run receipts

A receipt for every allow and deny.

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.

Dhavan Shah 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.

Recorded test12 Oct 2026
Limit decisions
50
Allow receipts
20
Deny receipts
30
Workers that retried
50
New reservations from retries
0

Decisions without a receipt0

Automated test on PostgreSQL, not customer traffic.

A cost alert rarely explains what caused it.

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 timeline you can read

    A clear path through the captured model calls, retries and limit decisions for a given run.

  • “Is this the same failure as last week?”

    Incidents you can review

    Recurring evidence can be grouped, assigned and reviewed, with the reason recorded.

  • “Did the fix change anything?”

    Comparison between real runs

    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?”

    Unknown stays unknown

    Missing or stale cost evidence is not silently changed to zero.

Explain a costly run without stitching logs together.

How it works: from alert to conclusion.

Run timelines and incident records use the same run context as Agent FinOps.

1 · Capture

One run, one context

Cost, model calls, retries and limit decisions are tied to the same run.

2 · Investigate

Go straight to the cause

Move from an alert to the costly step, retry or limit decision behind it.

3 · Record

Leave a conclusion behind

Ownership, status and reason are kept, so your team can review the decision later.

What an incident record looks like.

Illustrative incident evidence from a captured run.

workflow:report_generationrun_id:run_01J...incident:repeated_model_retryoccurrences:4cost_state:settledreview_status:needs_reviewcomparison:existing_runs_only

What is verified today

Incident workflows are verified locally with synthetic data. Customer production use requires separate verification.

Where the boundary is

Comparing existing runs is evidence, not causal proof or an automatic fix. Hosted evidence uses sample data.

Before you get in touch.

Who is this for?
Teams whose stack already holds the evidence, but whose engineers still rebuild the story by hand.
Does it replace my observability tools?
No. Run receipts accept OpenTelemetry traces, so they sit next to the observability you already use.
Are prompts sent to KazenAI?
Prompts and completions are not sent to KazenAI by default.
Can an incident expose a run someone can’t see?
No. Derived incidents and exports do not grant access to the underlying runs.
Is this a separate product?
No. Run receipts are part of Agent FinOps.
Who built it?
I did. Run receipts are the explanatory half of Agent FinOps: they connect the cost and spending decision to the calls, retries and incident record behind the run.
Can I see the product?
Book a margin review on kazenai.com. The hosted demo uses sample data.
How do we start?
Send a short description of the workflow and what is blocking it. I reply within one business day. The first 30 minute call is free, and there’s no obligation until you approve the Diagnostic.

Bring one expensive or failed workflow and the investigation gap you want to close.

The first 30 minute call is free, and I reply within one business day.

Book a free call

How engagements work · Back to Projects