We do the complicated stuff so you don't have to.
Here's the whole idea behind / foxlight: no terminal, no prompt engineering, nothing to configure. Just software that works the way it should. We want powerful AI and strong security to feel effortless, run on your own device, and stay fully under your control.
Most companies lean on giant AI models that need massive server farms to run. We'd rather train small, focused ones instead. Using a technique called LoRA, we can teach a lightweight model to be really good at one specific job — without the cost or bulk of a full-sized model. The training happens in the cloud, but running the model afterward happens right on your own device.
Because these models are small, we can swap them out while the app is running — no restarts, no downtime. That means your AI keeps getting better at its job without you ever noticing a hiccup.
Getting real work done with AI agents is usually messy — juggling multiple agents, splitting up tasks, keeping track of budgets and priorities. We handle all of that behind the scenes so you never have to touch it.
In Obscura, for example, you just type what you want — something like "put together a launch across all my platforms" — and a group of AI agents works out the steps and gets it done on their own. You set the goal. We handle the rest.