HoodAI reads your on-chain positions and answers in plain English — while the AI model, your keys, and your data never leave your machine. No servers. No accounts. No surveillance.
HoodAI's future runs on everyday devices. Share your spare GPU now, and when the compute network goes live you'll earn tokens for serving its AI.
Most "AI finance" tools ship your portfolio to a company's servers. HoodAI flips it: the model runs inside your browser, on your own GPU.
Robinhood Wallet, MetaMask, or paste any address. It shares only your public address — read-only. Your keys stay in your wallet.
Pick a model for your device. It downloads into your browser and is cached — then runs fully offline on your GPU.
“What's my risk?” “Explain this token.” “Am I near liquidation?” Answers grounded in your live positions — computed locally.
The model weights, your wallet data, and every prompt stay local. There's no backend to leak, subpoena, or get hacked.
Tokenized stocks, stablecoins, DeFi, and memecoin dust — untangled into a clear read on what you own and what it's worth.
Concentration, stablecoin ratio, gas checks, and (soon) background alerts when a position moves or a loan nears liquidation.
Read-only today. When execution lands, HoodAI proposes — your wallet signs. Or grant a scoped, revocable permission. Never your seed.
Running the model on your own device isn't just about privacy — it changes the economics and points at where AI is going.
Cloud AI bills you per message. HoodAI runs on hardware you already own — $0 per query, no subscription, no API bill. Ask as much as you want.
No rate limits, no “try again later,” no company deciding your access. Once the model's cached it even works offline.
On-device models are now good enough to be genuinely useful — and only getting better. Local AI is where this is heading; HoodAI is built for it.
Bigger models are smarter but need more memory. Here's what each needs — the app auto-suggests one based on your device.
| Model | Best for | Download | RAM / VRAM needed | |
|---|---|---|---|---|
| Llama 3.2 1B | Phones, low-end | 0.9 GB | ~4 GB | |
| Qwen2.5 1.5B | Phones+ | 1.1 GB | ~6 GB | |
| Llama 3.2 3B | Laptops | 2.0 GB | ~8 GB | |
| Qwen2.5 3B | Laptops · best tool-use | 2.0 GB | ~8 GB | Recommended |
| Qwen2.5 7B | Strong desktops | 4.7 GB | ~16 GB |
Local read-only today. Next: a native app with scoped execution and a token-powered marketplace of strategies. Then a network where you rent out your GPU and earn tokens for serving compute.
No install, no account. Just a link that runs a private copilot in your browser.