Launching today
Local is a macOS app that runs fully private AI directly on your own machine. We built it for completely frictionless setup. It tunes itself to your hardware, so chat, coding agents and meeting notes run up to 5.4x faster on the same hardware, without your data ever leaving the device. No cloud. No accounts. No cost. "Office Mode" let's you run AI on the fastest machine in the office and every laptop can connect to it.
BaseRT
Product Hunt
Congrats on the launch. I'm curious if you ran tests on different hardware specifications and if you noticed an ideal setup for certain models and what the bare minimum requirements are.
Cool! Does it only work on Apple Silicon processors? I have a 2020 MacBook Air, the last model with an Intel chip. Would it work on mine?
The hardware auto-tuning is the part I want to verify. I have tried a few local inference setups that claimed to optimize themselves but still needed manual model config to get decent throughput on an M3 Pro. If Local handles that without any manual tuning, what context window does it target by default, and can I override it for longer coding sessions without undoing the hardware fit?
BaseRT
If you're curious about the engine that makes Local the fastest AI inference app on Apple silicon, check out BaseRT and the research behind it: https://www.basecompute.co/research
BetterClaw
LLM costs can get messy really quickly. Having a smarter way to balance price and performance feels like a problem worth solving.