Your project never leaves our building
A $5,000 one-time setup that moves all AI-assisted work on your project onto self-hosted models running on hardware in our Mukwonago office. No third-party AI service ever sees your data.
What you are buying
We build with AI on every project. By default that means Claude Code from Anthropic, which is the best tool for the job and what we reach for unless you tell us otherwise.
Some businesses cannot send their data to a third-party AI service. Not because they distrust the vendor, but because a contract, a regulator, or a customer says no. This option exists for them. We provision a dedicated machine in our office, load open-weight models onto it, and run your entire engagement there.
What stays inside the building
Your data, your code, your configuration, and every prompt written about your project. The model runs on our hardware. Nothing about your engagement is sent to an external AI API.
What you still get
The same consultants, the same delivery process, and the same senior review on every change. The tooling underneath changes; the accountability does not.
Who this is for
- Defense and aerospace suppliers working under ITAR or CMMC obligations
- Healthcare organizations handling PHI
- Law firms and accountants holding privileged client material
- Manufacturers with process data or drawings they treat as trade secrets
- Any business whose customer contracts prohibit third-party AI processing
How it works
- We provision a dedicated machine. A GPU workstation in our Mukwonago office, set aside for your engagement and not shared with other clients.
- We load open-weight models onto it. Meta's Llama family and comparable open models, downloaded once and run entirely on that hardware.
- Your project data goes on that box and stays there. Your Odoo database copy, your exports, your documents.
- We work the engagement from that environment. Configuration, migration scripts, custom modules, and reports are all produced there.
- You get a written record. What was provisioned, what ran where, and what was destroyed at the end. Something you can hand to an auditor.
The honest tradeoff
Open models are good. They are not yet Claude.
On the hardest agentic work, self-hosted open-weight models still trail the best commercial models. In practice that means some tasks take us longer, and a few are better done by hand. We will tell you when that happens rather than quietly burning your hours.
It costs more, not less
You are paying $5,000 for privacy, not for speed. If your data has no restriction on it, buy the standard engagement and put the $5,000 into more hours instead. We will say so on the call.
We would rather lose this sale than have you buy protection you do not need.
What it costs
$5,000 one-time setup. That covers provisioning the machine, standing up the models, isolating the environment, and producing the written record at the end.
Your consulting hours are billed at the normal rates, from $65 an hour down to $50 on the larger packs. The setup fee is charged once per engagement, not per month.
Questions
Is the machine air-gapped?
The model runs locally with no external AI calls. Full physical air-gapping is available if your requirements demand it, and it changes how we work day to day, so raise it early and we will scope it properly.
What happens to my data when the project ends?
It is destroyed and you get written confirmation of what was removed and when. If you need the environment kept warm for a future phase, we can arrange that instead.
Can we host it on our hardware instead of yours?
Often yes, and for some organizations that is the better answer. It is a different scope, so bring it up on the first call.
What do you use if we do not buy this?
Claude Code from Anthropic, which is our default tooling on every other engagement.
Have data that cannot leave?
Bring the restriction you are working under. We will tell you plainly whether this option solves it and whether you need it at all.
Schedule a Consultation