PearPie

PearPie is a desktop and mobile chat application that lets you run open-source language models locally on your own hardware at no cost, or route requests to frontier models through a European server network using a credit system. The workflow requires no account creation: you open the app, your device generates a cryptographic key, and you pick a model per conversation. You can switch models mid-chat. Conversations are stored only on your device and sync to your other devices over direct peer-to-peer connections.

The defining architectural choice is what PearPie removes from the standard AI chat stack. There is no central server receiving or retaining your conversations, no user profile being assembled, and — for premium requests — the vendor states your message is discarded after the reply is returned and is not used for model training. All processing, local or network-based, happens on European infrastructure. A peer-to-peer GPU sharing feature lets you access a trusted contact’s hardware over a direct secure connection, with no server in the middle.

PearPie fits a specific profile: individuals or small teams where privacy is the primary concern and a conversational interface covers the full use case. It does not expose an API, so it cannot be wired into a pipeline, an automation, or an external application. It has no agent capability — there are no tools, no task loops, no actions the model can take beyond generating a reply. Teams that start here and later need those capabilities will rebuild elsewhere. The free tier is limited to models that run on your own hardware; access to frontier or larger hosted models requires credits, which is a paid-only feature.