gitmyhub

sage

Shell ★ 306 updated 1d ago

A desktop app for running AI models fully offline on your own computer, with a Rust-based inference engine, a built-in editor and terminal, and OS-level sandboxing for agent actions.

Rustsetup: easycomplexity 3/5

Sage is a desktop application for running AI models entirely on your own computer, without sending anything to a remote server. The team built their own inference engine from scratch in Rust rather than wrapping an existing tool. The README describes the goal as keeping AI performant and local, with nothing leaving your machine and no accounts or network connection required after installation.

Beyond just chatting with a model, Sage includes an editor, an integrated terminal, file viewers, and an IDE-like environment, so you can work with AI across the surfaces where you actually build things. A notable feature is the security model: all agent actions and tool calls run inside a kernel-level sandbox enforced by the operating system, not by the AI model itself. This means the AI can only access files and resources you have explicitly permitted, and that boundary cannot be overridden through prompting.

The app is available for macOS (Apple Silicon and Intel), Windows (x86-64), and Linux (AppImage and .deb formats). Installation is described as one click: download, open, and it runs. There is no Docker, no servers to configure, and no config files to edit. The app checks for updates on each launch, downloads them in the background, and offers a restart prompt to apply them.

This GitHub repository exists only to distribute the compiled application and publish release notes. The inference engine and agent runtime are proprietary and ship as binaries only; no source code is included here. The application is free to install and use but may not be redistributed or reverse-engineered. It is built by Conifer Solutions.

Where it fits