Ten people,one box, no seats
A team's whole AI day — chat, code, automations — served concurrently from one machine with no per-seat anything.
Nothing your team types leaves the building — and nobody audits your seat count.
The day, step by step
Each step names the tool and the model doing the work — all from the verified catalog.
- All dayOpen WebUI · gpt-oss-120b
The shared front door
Everyone on the team chats against the box with their own account and history — the measured concurrency model serves 286 tok/s aggregate.
- All dayContinue / Cline · Laguna S 2.1
Engineers on the same box
Coding agents and autocomplete against the local endpoint; the flagship on tap for the hard problems.
- Hourlyn8n · Qwen3.5 122B
Ops automations
Support inbox triage, CRM hygiene, report drafts — automation flows calling the box as a node.
- FridaysOpenClaw · DeepSeek V4 Flash
The weekly digest
What shipped, what stalled and what the support queue is saying — assembled from the team's own tools, distributed internally.
The stack, pre-installed
- gpt-oss-120b
- Laguna S 2.1
- Qwen3.5 122B
- Open WebUI
- Continue
- n8n
- OpenClaw
The honest caveat
Ten concurrent chat users is measured territory; ten simultaneous heavy agent loops is not — that is what the second box is for, and the configurator says so.
This arrives working — models loaded, agents wired, your team walked through it.
Configure yours