Built by ideius. Guarded by Rover.

Why local-first

Cloud AI is convenient right up until it isn't. Your data leaves your machine, the bill scales with how much you use it, and the whole thing stops working on a plane or when an API has a bad day. Apple Silicon changed the maths. A modern Mac has enough unified memory to run real models locally and fast: chat, images, video, voice, vision, perception, search. Bedrova exists so you can actually do that without stitching six apps together. One app, one API, every model type, nothing leaving your Mac.

We compose the best open work

We don't claim secret inference tech. Bedrova stands on osaurus, MLX, mlx-lm, mflux, mlx-audio, llama.cpp and Core ML, with our engineering tying them into one reliable runtime. We're honest about what's under the hood, and honest about the boundary: Bedrova runs every kind of model, drawn from what MLX, Core ML, GGUF and our documented provider packs can actually run on a Mac. Not literally every model. When something won't load, Bedrova says so and explains why.

The canine family

Rover is your loyal local companion: he fetches models, keeps the good ones close, and guards your data. When work needs to run on a schedule across a team, the pack grows up into Packwolf — agents and durable knowledge. And when a business wants it built for them, that's ideius. One universe, three rungs; use exactly the one you need.

Building in the open

We build in public — real benchmarks, honest trade-offs, no hype — narrated by Rover on the blog. When something is gated or unproven, we say so.