Two engines. Offline or Online —
you're in control.

hYYa ai is one privacy-first AI app with two engines you switch between anytime: hYYa Cloud — the online engine — connects to the world's most powerful AI models, and hYYa Vault — the local engine, out now in the hYYa ai Mac app (iOS and Android still in development) — runs entirely on your device with no internet required. Same app, same memories; you choose the engine, every time.

They are not two products to choose between — they are the two engines inside one hYYa app, and having both is the thing no other assistant does. Every other option is one or the other: a cloud service, or a local model runner, and moving between them means moving apps and leaving your history behind. hYYa Vault is the local engine — it runs entirely on your device, works offline, needs no account, and is free. hYYa Cloud is the online engine — it taps the world's most powerful AI models over the internet. The native apps carry BOTH: hYYa ai for Mac is out now on the Mac App Store, and iOS and Android are still in development. In a browser you get hYYa Cloud only, because a browser tab cannot run a local model. Switch engines anytime; one memory follows you across both.

On the web today, hYYa runs the Cloud engine, which needs an internet connection. The hYYa Vault engine — out now in the hYYa ai Mac app, with iOS and Android still in development — runs entirely on your device with no connection needed. No Wi-Fi, no mobile data — works on a plane, in a basement, anywhere.

The hYYa ai APP is free — download it and hYYa Vault runs entirely on your device, with no account and no payment, forever. hYYa CLOUD is the paid half: there is no free Cloud tier and no trial. To use it you choose one of two — a one-off credit pack from $20 (nothing recurring, nothing to cancel, and unspent credits are refundable within 14 days) or a Cloud Ultra subscription at $49.99/mo.

They carry exactly the same toolkit — the full-corpus DataLens, team lenses, watermark-free output, the frontier models (Claude Opus, Claude Fable, Grok) and unlimited everyday chat are on both — so you are not choosing features, you are choosing how to pay. On-Demand (from $20) is a one-off: buy a credit pack, spend it at your own pace, nothing recurring and nothing to cancel; credits never expire, and unspent ones are refundable within 14 days. Ultra ($49.99/mo) is the subscription: a larger allowance that renews every month, the cheaper advanced models at no credit cost, and Bring Your Own Key (BYOK) — the one capability On-Demand does not have.

Because the stores take a cut. Apple and Google charge a commission on everything sold through their stores, and that cost has to live somewhere — so the same hYYa Cloud plan can carry a different price tag in an app store than on hyya.ai. What you get never changes: the same features, the same credits, the same hYYa.

It is one account everywhere hYYa runs. Sign in on the web or in the Mac app — out now — and your plan, your credits and your chats are all there. The iPhone and Android apps — still in development — will use the very same login.

When it's time to top up or start a new plan period, buy wherever suits you that day — in the app or on the web — and everything lands on the same account. Your account follows you, not the store.

DataLens lets you chat with your own documents. Upload PDFs, contracts, reports, or research papers and ask questions across them, with citations — available on every tier, drawing from your allowance. Both paid tiers, On-Demand and Ultra alike, get the full version: whole-corpus reasoning across large sets with Claude Opus, OCR of scanned & bulk documents, and shared team lenses with up to 10 teammates.

Connections securely link apps like Gmail, Google Calendar, Outlook, Notion and Asana so hYYa can use your real data in chat — read-only, on every tier. You choose what to connect and can disconnect any app anytime from Settings. hYYa never sells your data or uses it to train AI.

Yes — Live Translate is a real-time voice interpreter: speak in one language and hYYa speaks it back in another, hands-free, across 14 languages — English, Arabic, French, Spanish, German, Italian, Portuguese, Turkish, Indonesian, Hindi, Urdu, Chinese, Japanese and Russian. It's part of Voice, available on every tier and metered from your monthly credit allowance.

Yes — hYYa Ultra supports Bring Your Own Key. Add your own Anthropic, Google, Groq, and other provider keys to use your own models and billing. Your keys are encrypted and never shared.

Yes — and it's now built in, no separate add-on. Open a Drawings lens and hYYa measures a real take-off straight off your construction drawings — lengths, areas and equipment/block counts computed from the drawing's geometry (never by eye). On a native AutoCAD DWG or DXF you also get per-layer wall centrelines (parallel faces collapsed, not doubled line-work) and exact block counts in real units with no calibration; a vector PDF just needs a quick two-click scale. One click exports the whole take-off to Excel/CSV, and your drawing never leaves hYYa. Curved line-work is measured at its true developed length — except SPLINEs, which hYYa does not measure: it counts them and says so on the take-off with a ⚠ Coverage line rather than estimating, so nothing is dropped silently. On the native hYYa Vault app the take-off runs fully on-device from vector PDF and AutoCAD DXF drawings; per-layer wall centrelines are hYYa Cloud only, and a binary .dwg is read by hYYa Cloud only — save it as DXF first if you are working offline. Freeing the data trapped in your invoices, purchase orders and forms into an exact Excel workbook is built in too. Every capability draws from your monthly credit allowance. Tired of chatbots that only talk? Experience hYYa ai that executes.

These are approximate and vary a lot with document size and answer length. FREE (starter allowance, per month) ≈ 10 images, or 7 Word/Excel exports, or 54 take-off exports, plus a daily allowance of everyday chat. A $20 ON-DEMAND pack ≈ 30 flagship images, or 45 Word/Excel exports, or 11 PowerPoint decks, or 327 take-off exports — and it is a balance, so it lasts until you spend it rather than resetting. ULTRA (per month) ≈ 91 images, 137 Word/Excel exports, 34 decks, 1,000 take-off exports, and full-corpus DataLens questions on top — that last one is the single most demanding thing hYYa does, and it is the one cost that is NOT flat: it scales with how much you have uploaded, and on which of the two DataLens models you choose. On a large set (600,000 tokens) one question is about 9,000 credits on the default Standard, or about 15,000 on Deep — so Ultra's month covers roughly 6 of them, or 3-4 on Deep; on a small set, many more. Everyday chat on the standard models doesn't touch your allowance at all on a paid plan, and on Ultra the advanced models don't either. Mix and match freely: the allowance is one pool, not per-feature buckets.

Your monthly credit allowance refreshes every month, so most people simply wait for the refresh — and if you run out every month, moving up a tier is better value per credit than topping up. If you just need to finish the month, one-time packs are available on every tier: 18,000 credits for $20, 45,000 for $50, or 90,000 for $100 — all at the same rate. A pack is a one-time purchase, never a subscription, and purchased credits are yours to keep — they never expire.

Each hYYa coupon works on one specific tier and one billing cycle. Check your coupon for the tier (On-Demand or Ultra), then choose that plan before checkout. At the secure Polar checkout, add your promotion code and confirm the discount appears before you finish.

You can always sign in to hYYa, even after a paid subscription ends — your account simply moves to having no plan tier. Your conversations and DataLenses are kept for 60 days, so you can pick up exactly where you left off if you resubscribe, and hYYa emails you reminders during that window. After 60 days without an active paid plan, that document data is permanently deleted. You can resubscribe anytime, or delete your data sooner from Settings.

hYYa Cloud conversations are never used to train AI models and are not sold or shared. The hYYa Vault engine (in the Mac app today) stores all conversations exclusively on your device — nothing sent to any server. hYYa ai is built privacy-first from the ground up.

hYYa runs on the web at hyya.ai, where you get the Cloud engine — a browser tab cannot run a local model, so that is a limit of the browser rather than a plan. hYYa ai for Mac is out now on the Mac App Store — free, no account needed, Apple silicon and macOS 26 or later — and it carries BOTH engines in one app: hYYa Cloud online and hYYa Vault entirely on your device. iOS and Android are still in development and will carry both as well. One hYYa Cloud plan covers every platform, and hYYa Vault costs nothing on any of them.

hYYa Cloud gives you access to the world's most capable AI models — including Claude, Gemini, Grok, Kimi, GLM and Qwen. It also supports web search and DataLens — chat with your own uploaded documents — all from the same app.

An account is needed to use hYYa Cloud — sign in with just your email, no password required. Using hYYa Cloud needs a credit pack or a subscription; the hYYa ai app and its on-device hYYa Vault engine are free, and the freier needs no card. The hYYa Vault engine — in the Mac app today — doesn't require an account.