← All apps
OPENPOCKET / PRIVATE AIAvailable now

A little intelligence.
All yours.

Private AI for iPhone and iPad. Discover downloadable language models, keep conversations on your device, and bring your own documents into the conversation.

Available on the App Store · iPhone & iPad · iOS 17+

On your device
A model that feels like yours.Explore models and choose one that suits your device.Illustrated walkthrough based on the app. Example conversation; timing is illustrative.
A LOOK INSIDE

Your models.
Your train of thought.

A native app built around a simple idea: useful AI, with the conversation kept on your device.

Find your model

Explore downloadable models with guidance based on your device’s memory and storage. Choose and manage the models you want to keep.

Keep the conversation

Chat with a local model, then search, rename and revisit saved conversations. No account or cloud conversation sync.

Bring your own words and images

Import text, Markdown or text-based PDFs and review the text before attaching it. Ask about a selected photo with a compatible vision model.

Listen your way

Have responses read aloud with available system voices, with controls for speaking rate, pause and playback.

BUILT AROUND LOCAL AI

Room to think.
On your own device.

Privacy policy · Support & safe use

OpenPocket runs downloaded models locally. Messages, imported document text, supported image attachments and responses are processed on your device, with no cloud AI fallback.

An internet connection is needed to download models. Conversations and models are stored locally; you choose what to share or delete.

No account. No analytics. No cloud conversation sync.
READY WHEN YOU ARE

Your next idea.
Already in your pocket.

Download OpenPocket, find a model that fits your device, and start a conversation of your own.

Download on the App Store
Available now
Made for
iPhone & iPad
Requires
iOS / iPadOS 17 or later
After model download
Chat offline

Model availability and performance depend on your device. Photo understanding requires a compatible vision model. Scanned PDFs are not supported. AI responses can be inaccurate; check important information.