Actual is a software platform that helps people run open AI models on computers they already own. Designed for private, locally controlled inference, it turns supported macOS, Linux, and Windows devices into a connected computing setup for AI. Users can install Actual on multiple computers and link them through an encrypted relay, choosing between on-device inference through a local endpoint and remote access through the relay.

The platform is intended to give users control over their keys, hardware, and computing environment. Actual also offers an optional way to contribute idle machine capacity to a shared pool and receive compensation for those contributions. Getting started involves installing the software and authorizing the device through Actual’s website; account registration is available for users who need it.

The service includes model exploration and documentation, with setup information covering installation and local and relay inference APIs. By using existing hardware rather than requiring a dedicated AI computer, Actual aims to make private AI inference accessible across a user’s own devices