The questions we get right before people buy.

Specifically: after the 7-day install, we identify a waste pattern where fixing it (and only that fix — no hypothetical compounding) would reduce your AI spend by at least $50/month at your current call volume.

The dollar number is projected from what we actually measured, not a guess. Your report shows the math. If the total across all findings is under $50/month, you get a full refund — no questions, no argument.

Yes. If we find real waste but it's below $50/month, we refund the difference prorated at $2/month — so $30/month of findings refunds $40 of the $100. That way you only pay in proportion to the value you got, and we don't take $100 for a report that didn't clear our own bar.

No. The report tells you what it would save if you implemented the fix. We don't take credit for savings you didn't actually realize. You're paying for the diagnosis — you decide whether to ship the cure.

Yes, but with caveats. If your AI spending is hidden inside a tool you don't control (e.g., you pay ChatGPT Team $25/month and that's your total AI spend), there's nothing for us to measure — the provider isn't billing you per-call.

If you or your team builds something that makes AI API calls (a chatbot, an assistant, an agent, a feature inside a product), that's where KostAI has something to measure. If you're unsure, just email and describe what you've got; we'll tell you honestly if it's a fit.

It works in AI assistants that can run shell commands on your project — Claude Code, Cursor, Windsurf, Warp, and some self-hosted agents. It does not work in plain Claude Desktop, the ChatGPT web app, or Gemini web, because those don't execute code on your machine.

If you're using one of the dev-agent tools, just paste the prompt. If you're using plain Claude or ChatGPT and want to try the free scan, email us and we'll walk you through it — or use the npm install kostai path directly if you have Node on your machine.

Yes. Enterprise installs are custom-scoped — multi-team setups, private network deployments, long-term monitoring. Prices start around $2,500 depending on scope. Email for a conversation.

KostAI is the focused one: it watches your AI calls for a week and tells you exactly what's wasting money. One-shot diagnostic. Clear dollar answer.

BrainOfBrains is the ongoing version: it leaves KostAI watching permanently, plus adds a handful of other "little brains" that keep an eye on different things (team signal, revenue, agent throughput). Same install, bigger scope.

If you just want your AI bill down: buy KostAI. If you want the always-on version with more watchers: buy BrainOfBrains. Same $100 either way; BrainOfBrains includes KostAI.

No. The install runs entirely on your infrastructure. We set it up on a video call and after that, you own the files. We never receive your event logs, your prompts, or your API responses.

The only thing we see is what you send in your email when you buy, and what you choose to paste back in the "day 7 report" reply. You're the one who decides what to share.

Nothing. The install is entirely on your machine — Python/Node scripts, launchd job, JSON files. It keeps running. You keep using it. There's no SaaS backend, no API dependency, no kill-switch.

Ten minutes on a video call. We screen-share, install on your machine, verify the first measurement comes through, and you're done. No overnight setup, no email back-and-forth.

Primary: macOS. Most customer stacks: Node.js or Python wrapping the Anthropic or OpenAI SDKs. If you're on Linux or using a different AI provider (Google, Mistral, local Ollama), it probably still works — email first and we'll tell you before you pay.

Yes. 30 days from install, full refund, no-questions, for any reason — even "I just didn't want this after all." We'd rather you keep your $100 than be unhappy with the install.