Your Own AI

Your Own AI is a free, open-source desktop app for macOS, Windows and Linux that runs AI models offline on your machine, with private signed memory, smart routing and optional paid online frontier models.

Your Own AI is a free, open-source desktop application that runs language models directly on the user's computer. With offline models, inference happens on local hardware, so conversations cannot leave the machine. The app is released under the AGPL license, which lets anyone inspect the code and confirm that claim instead of relying on a privacy policy. It targets people who want a private ChatGPT-style assistant for sensitive material such as contracts, medical records, code and personal characters. In practice the app handles several jobs in one place. Users can ask questions about attached documents and check answers against the exact source lines, open a project folder so the AI can read files, edit them and run commands with permission, and author characters with their own personalities and memories. Health use is supported by adding MedGemma, Google's medical model, to read lab results and scans locally. Once models are downloaded, chat, documents, characters and memory work with no connection at all. Online access is optional. Paid plans add frontier models and web search with cited sources through Flowsta's relay, which strips identity and does not store or log messages. Auto routing reads each message and picks a model based on difficulty, need for current information and hardware limits, with an offline-only mode where nothing leaves the device. Usage is billed per use against a monthly allowance, and local features keep working if a plan is cancelled. Underneath, every chat, project session and API request becomes a cryptographically signed transcript built on Holochain, stored encrypted on the machine under a key only the user holds. Memory the AI forms traces back to its source conversation. A local OpenAI-compatible endpoint lets editors, agents and scripts use the same models, memory and records. Compared with cloud assistants, it trades convenience and raw model size for control, while compared with plain local runners it adds routing, memory and auditability.