Four ways in, one runtime underneath.
Whichever surface you pick, the engine, the model store, and the scheduler are the same code. Try it in the browser first — no install. When you’re ready for native speed, sign in and grab a build.
Desktop app
The Tauri shell hosts the Conifer engine in-process. Models live on disk, the scheduler runs on your accelerator, and the chat UI is the same surface as the browser build — only faster. No telemetry, no cloud roundtrip.
Command line
A single static binary (`conifer`) for terminals, build pipelines, and headless boxes. Same engine as the desktop app — no daemon to keep alive. Stream tokens to stdout; pipe them anywhere.
One-line installer
Drops the CLI into ~/.conifer/bin and points your shell at it. The script is signed and reviewable — read it before you pipe it.
In your browser
No download, no account. Conifer's browser surface picks a model that fits your machine, downloads weights into IndexedDB, and runs inference on WebGPU. Slower than native, fine for kicking the tires.