ownAI

ownAI

The AI that belongs to you.

0 ratings
Free

Rating summary

Details

Released
August 21, 2026
Updated
September 30, 2026

Features

ownAI screenshot #1 for iPhone
ownAI screenshot #2 for iPhone
ownAI screenshot #3 for iPhone
ownAI screenshot #4 for iPhone
ownAI screenshot #5 for iPhone
🖼️Get Icon
Icons↘︎

About

ownAI is a personal AI that belongs to you — not to a platform. It remembers what you talk about and gets smarter over time. If it is missing a capability, it builds one itself. The more you use it, the more it becomes yours. YOUR DATA STAYS YOURS Conversations, memories, and settings are stored locally on your iPhone. When you use an external language model, only what is needed for a given request is transmitted — nothing else. No ads, no tracking, no data mining. ownAI is open source. ONE CONTINUOUS CONVERSATION No chat sessions to manage: ownAI is one flowing dialogue. A layered memory keeps what matters — recent context, summaries, and long-term memories you can browse and edit at any time. AN AI THAT EXTENDS ITSELF Ask for a new capability and ownAI builds it: small reusable helpers for everyday automation, or interactive mini-apps that run right inside the app, safely sandboxed. YOUR CHOICE OF MODEL - Entirely on your device: run an open-source model locally — private and able to work offline. ownAI checks your iPhone and recommends a model that fits it. - Via ownAI Credits: powerful models from leading providers with a single balance — you choose the region that processes your requests (EU, Switzerland, US). - Via your own API key: connect Anthropic, OpenAI, or your own Ollama server directly. ALWAYS WORKING FOR YOU Schedule recurring tasks and get a notification when they finish. IN YOUR LANGUAGE ownAI speaks English and German.
Show more

What's New in ownAI

0.7.0

September 30, 2026

Local models are now much better at using tools: - Tools and programs work far more reliably with local models, because the models now receive and write tool calls in the format they were trained on. If a call still goes wrong, the model is told so it can correct itself. - Local models start with the largest context your device can handle (up to 32k). You can pick a larger one by hand. - Local models now need at least a 16k context. 8k is only offered on devices that could not run any model otherwise. Fixes: - Models with recurrent layers (e.g. Qwen 3.5) now work from the second message on. - Long tool calls, such as creating a program, are no longer cut off and shown as raw text. - Gemma 4 models no longer run out of memory. - The answer after several tool calls is now always saved, so the AI keeps an accurate record of the conversation. - The AI no longer tries to use capabilities that were left out to fit a small context window.

More