About

Yes. Install the Infersec conduit on any Linux or macOS machine — mini PCs, desktops, Macbooks, rackmount servers. If it can run an inference engine like llama.cpp or vLLM, it can be a source.

Entirely on your own hardware. Infersec never hosts models — it handles routing, endpoint exposure, and policy while your machines do the inferencing. Nothing leaves your infrastructure.

Infersec detects offline sources automatically and reroutes traffic to healthy sources using fallback chains and priority rules. Endpoints stay available even when individual machines drop off.

That depends on your hardware and the model you choose. With a modern GPU or Apple Silicon, local models like Llama, Mistral, and Qwen deliver responses competitive with hosted services — often faster since there's no queue behind other tenants.

For many workloads, yes. Apple Silicon handles 7B–70B parameter models well, and Infersec's routing lets you chain multiple machines together. You won't match the largest proprietary models on every benchmark, but for most agent workflows, internal tools, and day-to-day inferencing, your own hardware is more than enough — and you keep full control of the data.