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 Agent FinOps

A hard spending limit for every customer’s AI.

KazenAI Agent FinOps checks every supported model call against the customer’s limit before it runs, and ties the cost to the customer and workflow behind it.

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
Concurrent agents
50
Customer budget
$5.00
Reserved per call
$0.25
Allowed
20
Denied before the provider call
30
Reserved in total
$5.00

Over budget$0.00

Automated test on PostgreSQL, not customer traffic.

Your provider bill shows how much was spent, not always who spent it.

Agent FinOps connects each supported model call to the customer and workflow behind it, then applies the agreed limit before the next call.

  • “Which customer or workflow drove this cost?”

    Cost by customer and workflow

    Cost records carry the customer and workflow your integration supplies, so you can see which customer facing work caused the cost.

  • “What stops the next call from going over?”

    A check before every supported call

    The agreed limit is checked before the next supported provider call. Over the limit, the call doesn’t run.

  • “What did we expect to spend, and what did we spend?”

    Reserved and actual cost kept apart

    Expected exposure and recorded actual cost stay separate in the record.

  • “Why was this run stopped?”

    One record for every decision

    Limit decisions and run receipts refer to the same run, so allow versus deny is visible and reviewable.

Make agent spend controllable before it hurts.

How it works: three steps on one workflow.

Adoption starts narrow. One existing workflow is evaluated before any wider rollout.

1 · Identify

Tag each supported call

Your integration adds the customer and workflow to each supported model call.

2 · Check

Check before the call

The agreed rule runs before the call. Where configured, expected exposure is reserved first.

3 · Settle

Settle and keep the receipt

Recorded actual cost settles the reservation, and the decision is preserved in a run receipt.

What a denied call looks like.

Illustrative evidence from a budget denial before a supported provider call.

HTTP 402 Payment Requiredavailable_budget_usd:0.20projected_spend_usd:0.50allowed:falseavoided_cost_usd:0.50provider_call:not executedevent:budget_reservation_denied

What is supported today

OpenAI Chat Completions and Anthropic Messages, sync and async, with or without streaming, verified against the live provider APIs. Other paths require separate verification.

Where the boundary is

The hosted demo uses sample data. Customer production deployment, shared policy behavior and your safe response after a denial require separate verification.

Before you get in touch.

Who is this for?
Teams that already operate customer facing AI workflows and need to know which customer or feature drove cost, and where a run crossed its limit.
Do I need this if I already have a gateway?
Not always. It’s most useful when your configured provider, gateway and observability stack still leaves a specific gap.
What happens when a call is denied?
The provider call doesn’t run, and the decision is preserved in a run receipt. You define what your workflow does next.
What does a limit cover?
Limits are checked before model calls on supported paths. They don’t cover your own tools, paid APIs or compute.
Who built it?
I did. Agent FinOps is the preventive half of KazenAI’s current product: attribute the cost, apply an agreed limit, and preserve the decision in a run receipt.
Do I have to buy KazenAI to work with you?
No. Clients can use it; it’s never required. Under the engagement agreement, what I build for you is yours and KazenAI stays mine.
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 customer facing AI workflow and the cost or control problem you want to solve.

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

Book a free call

How engagements work · Back to Projects