Open-source personal agenton your phone

Yes. Bring a key from a supported model provider, such as OpenAI, Anthropic, Google Gemini, or DeepSeek, and choose your model. Model usage and pricing follow your provider’s terms.

There’s no Anan server to set up: the agent runs on your phone and connects directly to the model service you configure. Your saved history is local, but cloud model requests need an internet connection.

Conversations, memories, and task records are stored locally; API keys are kept in the system Keychain. Relevant messages, attachments, memories, and tool results can be sent to your selected model provider when you use it. Our privacy page explains the details.

Anan works through supported phone tools, such as Calendar and Reminders, within system permissions. It cannot freely control every app. Writes normally ask for confirmation; you can manage saved allowances and automation settings.

You can schedule read-only work, such as a daily calendar brief, and review its results. Execution is best-effort and depends on iOS scheduling, connectivity, and available permissions. Tasks may run late or be missed; they are not a guaranteed alarm service.

It’s a planned way to connect your own cloud workspace for writing and running code, using your sandbox service credentials. E2B, Daytona, and Modal are shown as proposed providers. This feature is not implemented yet, and the website only previews its setup.

Anan is currently being developed for iOS. Download information and the GitHub repository will be shared here when ready. The open-source license is still being decided, and Android is not available yet.