On-premise first
Your hardware. Nothing leaves unless you say so.
The Beelab is on-premise first. That’s the whole point. Your data lives on your hardware, your AI runs on your hardware, and your data stays on the rack; control-plane traffic for Tailscale, Cloudflare Tunnel, and similar services is outbound-only and minimal by design. But Beelab isn’t trapped in the rack either. Optional burst paths route to any major cloud (AWS, Google Cloud, Azure, or your provider of choice) and any S3-compatible offsite (Backblaze B2 default, yours to swap), through an OpenAI-compatible gateway (default, yours to swap) and the Model Context Protocol.
Built for operators
Host your own apps, your own AI, your own pipeline.
This is for the people who actually want to use it. People who want to host their own websites end to end. People who want to build full-stack applications with AI integration, their own database, and a CI/CD pipeline they control from commit to production. People who want a real sandbox to break things in, an isolated DMZ for anything they put in front of the public internet, and the option to grow the rack from 1 node to N on the day they outgrow the first one. Plug a validated node in, follow the guided cluster-add wizard, the new node joins after a verified handshake. People who want to self-host their private services on hardware they own. People who want to train their own local AI on their own data, and run a private RAG (Retrieval-Augmented Generation) layer over their own documents.
Default-local RAG
The lookup runs on hardware you control.
The RAG layer is integrated on the rack itself, with guardrails and rules built in. Default-local RAG. Opt-in cloud burst (Bedrock, Vertex, Azure OpenAI, Anthropic, or your provider). Want to find a file you wrote three months ago? Ask the local assistant. The lookup runs on hardware you control, with prompts staying inside your network unless you opt into a cloud provider.
