A private, on-device AI chatbot — inference runs locally and your data stays on the device.

PocketMind is an AI chatbot that runs a local language model directly on your phone, so inference happens entirely on-device and your conversations never leave it. It supports two inference backends — GGUF (llama.cpp) and MediaPipe (.task) — and ships with a built-in model catalog: open-license models download in-app with a single tap, while gated models are guided through the official Hugging Face license flow. You can keep several models on hand and switch quickly, organize work across multiple persistent chat sessions with search and time grouping, and act on individual messages (copy, regenerate, delete). It works fully offline — the network is used only to download models you choose. The App is free and ad-supported, with an optional in-app purchase to remove ads.