Stop deciding the same thing twice.
In a small business the decision gets made by whoever is in the room that day — and months later nobody remembers why. Kirat lets Claude put the question to ChatGPT and Gemini as well. You get three independent positions, the majority, and the objection that lost, written down at the moment you decide.
It is not advice. It is structure.
Big companies don't decide better. They decide with more people in the room.
What a large organization has is not more intelligence. It is more perspectives before the call and a record after it. That structure is what a small business lacks, and Kirat gives you part of it.
The decision gets made by whoever argues hardest, or by you alone at the end of a long day. Nobody had time to prepare the counter-case.
There is no board and no advisors. Outside advice has to learn your business before it can disagree with you.
Months later, the same question is back on the table, because nobody wrote down what was considered the first time.
When a call turns out wrong, there is no record of the alternative that was on the table. So the same mistake stays available.
It pays off with decision volume, not company size.
A council is worth it when you have many decisions to make. If your month holds a handful, Kirat is not for you: two browser tabs are free and serve you just as well. Ask ChatGPT and Gemini yourself and compare.
FIG. 1 — The test that matters more than company size.
With volume, two tabs turn into copying and pasting, and nothing is written down. Kirat lets Claude put the question to ChatGPT and Gemini, and the result keeps each position, the vote and the objection that lost. The council doesn't decide for you. It puts the objection in front of you and keeps it in writing — and you overrule it when you know something the models don't.
What this does not do
- —It is not advice. Not legal, not tax, not financial. For those you need a professional who is accountable for the advice. Kirat gives you positions, not advice.
- —It knows nothing about your business. It knows what is in your question. Vague context in, confident nonsense out. Different model families fail differently, but three models can still be confidently wrong together.
- —It does not know your market, your customers or your numbers. On anything that turns on those, a person who does know them will beat it — and should.
- —It does not keep your data in the company. Everything you give Kirat goes directly to OpenAI and Google, under your accounts and their terms — from your computer, or in Claude Code cloud sessions from your own cloud environment through Anthropic's network proxy for that environment. If your data may not go there, this is not your product.
- —It needs a setup. Someone in your company uses Claude Desktop on macOS or Windows, or Claude Code, and creates two API keys with billing enabled — one at OpenAI, one at Google. That someone does not have to be a developer. Claude Code cloud sessions need a one-time setup of the cloud environment and come with limits: everyone who uses the environment can read its environment variables, including your keys; Kirat's daily budget counts per session, so set spending limits at OpenAI and Google; background jobs and recent results end with the session; and the environment must allow two hosts,
api.openai.comand the Kirat download host. If the setup fails, the session still starts, just without Kirat, and the failed state can stay cached for new sessions until you edit the setup script. Kirat does not run in chat in the browser or the mobile apps, or in Cowork tasks that run in the cloud (where it runs, in full). If your organization restricts Desktop extensions to an allowlist, an admin has to step in; networks on your own computer that require an HTTP proxy are not supported.
What setup involves.
There is no service to sign up for with us and no server for anyone to run. You buy once, download one ZIP and install the extension in Claude Desktop. If you decide it isn't for you, there is no subscription to cancel and no account with us to close.
FIG. 2 — Bill of materials. Claude Desktop warns that extensions installed from a file are not verified by Anthropic. That is expected.
You don't need a developer, and there is no server to run.
If you can install a Claude extension and create two API keys, you can use Kirat. Claude Desktop ships its own Node.js, so there is nothing else to install.
If you wonder whether software written by an AI agent can be taken seriously: Kirat itself is 11,351 lines of TypeScript, source and tests, in 38 commits — every commit written by an AI agent (Claude), none by hand, with 256 automated tests that run without network access.
What using it does require is recognizing when the answer coming back is wrong. That is a business skill, not a coding skill, and if you run a business you already have it.
Before you buy.
What kinds of decisions is this actually good for?
Reversible ones with more than one defensible answer, where you would want the counter-case in writing: pricing changes, whether to take on a client, build or buy, which supplier, how to structure an offer.
Bad fits: anything legal, tax or financial where you need someone accountable for the advice; anything that turns on numbers only you have, unless you put those numbers into the question; and anything you have already decided and simply want confirmed.
What does it cost to run?
Kirat itself is a one-time purchase — no subscription, no seats. Running it costs API fees, paid directly to OpenAI and Google: $0.02 to $0.15 per council decision in our measured runs, depending on effort tier and input size. Your monthly bill depends on how many decisions you send through it.
The budget guard reserves the worst-case cost before every call and blocks calls that could exceed the daily budget — $5 per UTC day by default, adjustable. It is best-effort, so also set spending limits at the providers.
Is our company data safe?
Kirat has no server, no account with us and no telemetry. We never see your keys or your data. But everything you give it (questions, context, code, images) goes directly to OpenAI and Google, under your accounts and their terms — from your computer, or in Claude Code cloud sessions from your own cloud environment through Anthropic's network proxy for that environment.
On your computer, Kirat keeps a .kirat folder in your home folder: the daily cost counter, the cost log, the last 20 results (they contain your questions and answers), the log of delegated decisions and generated images. You can delete it any time.
Practically: don't paste customer records, personnel files or anything you wouldn't send to OpenAI or Google today. If that rules out the decisions you care about, Kirat is not for you.
What if the council is wrong?
It can be, and sometimes will be. It is three opinions, not an oracle.
What it does give you is the argument you would otherwise never have heard, in writing. Read the dissent first; that is where the value sits.
What if we try it and it isn't for us?
There is no subscription to cancel, no seats to release and no account with us. Remove the extension, and delete the .kirat folder in your home folder if you want its local data gone. The API keys are your own accounts at OpenAI and Google; what you already sent there is under their terms.
Refunds: see Terms.
Who in the company should run it?
Whoever already uses Claude Desktop or Claude Code for real work — often one curious person rather than a job title.
A license covers one person, on any number of that person's own devices and own cloud environments. Others can bring their questions to that person; anyone who wants to run Kirat themselves needs their own license — including in a shared cloud environment.
One product for everyone: one purchase, one download with the Claude Desktop extension and the Claude Code plugin. Pricing and details are on the main page.