How it works
Three ways to chat, one app
PearPie gives you a choice in where the AI actually runs. Most of the time you don't have to think about it. But when you want to, here's what's happening behind the scenes.
The starting point: no account, no central database
The first time you open PearPie, the app generates a small cryptographic key on your computer. That key is your identity inside PearPie. It's how your other devices recognise you, and how PearPie's network knows it's really you when you connect. There is no email, no password, no profile, and no record of who you are anywhere on a PearPie server.
Your conversations get the same treatment. They're saved to a file on your own device, not uploaded to a backup. If you wipe the app, those conversations are gone. There's no copy we'd hold for you. That's the trade-off behind everything that follows: PearPie can't accidentally hand over your conversations to anyone if it doesn't have them in the first place.
Way 1 · Local
The model runs on your computer
PearPie comes with a small library of open-source AI models you can download with one click. Once a model is on your computer, your messages are processed right there. Your conversation never leaves the machine, and there's no internet round-trip happening when you press send.
For most people this is the most private option, because nobody but you can possibly see the conversation. It's also free and works offline. The trade-off is that bigger models need more powerful hardware. PearPie ships sizes that work on regular laptops and lets you scale up if you have something heftier.
Way 2 · Peer-to-peer
Your devices (and friends' devices) talk directly
PearPie has a peer-to-peer mode that lets your devices share AI work with each other over a direct, encrypted connection. There's no server in the middle storing or relaying anything. Two practical things this enables:
Sync between your own devices. Start a chat on your laptop, pick it up on your phone. Your conversations move between your devices over a private link, not via a cloud backup.
Use your home computer's GPU from your phone. If you have a powerful machine at home, your phone can borrow its brain when you're out. Same chat experience, much bigger model. Or share that link with someone you trust who'd benefit.
Peer-to-peer is like passing a note directly to someone in the room, instead of mailing it through a sorting facility. The note still gets there, but no third party touches it on the way.
Way 3 · The PearPie Network
Premium models, on European servers
The PearPie Network is for models you can't or don't want to host yourself: closed-source models like Claude (which can only be accessed through a hosted provider) and larger open-source builds of DeepSeek or Mistral that aren't practical to run on personal hardware. Underneath, the Network uses the same peer-to-peer mechanism as connections between your own devices: PearPie joins your network as a private peer running those models on EU infrastructure, under data processing agreements with the underlying providers. PearPie itself doesn't store the request. In some cases the underlying provider may briefly retain a request purely for abuse prevention before automatically deleting it; nothing is permanently stored by any third-party provider.
You only use this path when you choose a premium model. It costs credits, it runs in the EU, and PearPie still doesn't link any of it to a personal profile. Your identity remains your device key. The only exception: if you've purchased credits, your key is linked to a payment record at our payment provider (Stripe).
Premium models run on European data centres of large infrastructure providers (for example Claude is currently accessed through AWS in Europe). PearPie has data processing agreements with each one and acts as the privacy layer: nothing identifying you is sent along with your message, PearPie doesn't store the request, and any brief provider-side retention is for abuse prevention only. Over time, more of this infrastructure will move to PearPie-operated European hardware.
So which one runs when?
You pick the model in the app and PearPie figures out the rest. If you've chosen a local model, your message stays on your device. If you've chosen a peer model (your own desktop, a friend's machine), it goes over a direct encrypted link. If you've chosen a premium model, it goes to an EU data centre via the PearPie Network and the response comes back. Same chat window, different paths underneath.
Local and peer-to-peer are always free. The PearPie Network uses credits, because the GPUs running those big models cost money. See pricing →