The Offer
Own Your AI.
We'll Carry the Toolbag.
Every lane ends the same place: your hardware, your data, your kill switch. The only question is how much of the work you want to own.
What You Get on Day One.
Not a roadmap. Not a pilot phase. A checklist — where every line is something you can physically touch before we leave the building.
- A machine, configured and running, at handover. Pre-configured and ready on day one. You watch it answer before we pack up.
- Documented architecture your own people can maintain. Written for your team, not for us. If we vanished, the documentation wouldn't.
- A named internal steward, trained by us. One of your people learns to run it — because the goal is your capability, not our billable hours.
- A written agent operating agreement. Agents propose, humans decide — by architecture, not by policy. Not a promise a clever prompt can override; it's how the system is built.
- An audit log. What the system did, when, and on whose authority — recorded where you can read it.
- A kill switch. Yours. Physical. Ours to install, never to hold.
Every line above is a deliverable, not a description. If it can't be pointed at during handover, it doesn't belong on this list.
The Sovereignty Spectrum.
Not good-better-best. The question is simpler: how much of the work do you want to own? Every lane ends in the same place — you own the hardware, the data, and the relationship. The lanes only differ in who carries the toolbag.
- Scoped remote sessions: from boxed hardware or an existing home-server appliance to a working, documented private AI
- The field guide — the paid runbook you keep
- Governance templates: agent-constitution starter, audit-log schema, kill-switch pattern
- Cheap enough to buy without a committee
- Installed hardware, running at handover
- Documented architecture your people can maintain
- A named internal steward, trained by us
- Written agent operating agreement + audit log + kill switch
- 90 days of included care — then you choose: run it yourself (you can; that's the point) or roll into stewardship
- Monitoring, updates, model refreshes
- Nightly integrity discipline — the same one this site runs under
- Quarterly upgrade-path and obsolescence review, in writing, on a calendar — find another vendor who prints that
- Kill switch always in your hand
- Seats capped at what one steward can verifiably serve; the cap publishes here before launch
"MRC is the product. Hardware is the activation fee."
Pricing doctrine, stated plainlyWhat happens next is short: every path on this page leads to one 30-minute conversation. One qualifying question — build it, or have it built? Guided can close the same day. Built gets a one-page fixed-price scope within 48 hours. No proposals-as-theater, no discovery phases. We grow by referral and by record, not by ads.
How an Engagement Ends.
Most vendors design the engagement to never end. Ours is designed to end well — with you more capable than when it started. We win when you need us less.
"We don't want to own their data. We want to steward their stack."
Non-negotiable — every managed client owns their hardware and their dataAt handover, the following artifacts change hands — by name, not by gesture:
- Architecture documentation — the full map of what runs where and why, written for your team.
- Runbooks — the exact procedures for the things that recur: updates, restarts, backups, checks.
- The agent operating agreement — the written contract your AI works under. Agents propose; you decide.
- The audit log — the running record of what the system did and on whose authority.
- The kill switch — installed, tested in front of you, and left in your hand.
- A trained steward — one of your own people, able to run the stack without us on the phone.
If we disappeared tomorrow
Your system keeps running, because it never depended on anything on our side. Your documentation doesn't expire. Your steward already runs the routine. Your data was never anywhere but your building. That's the continuity plan — not a vendor's promise to be around forever, but an architecture where it doesn't matter.
Asked, Answered.
Won't the hardware be obsolete in two years?
Some of it, eventually — that's hardware. The difference is we put it in writing: every stewarded seat gets a quarterly upgrade-path and obsolescence review, on the calendar, that tells you when replacing a component is worth it and — just as often — when it isn't. One hardware investment replaces an open-ended subscription; the review is how we make sure it stays a good one. We haven't found another vendor who prints an obsolescence promise. That's why ours is in the contract, not the brochure.
What happens to my data if we part ways?
You keep everything — because there was never a copy anywhere else. Your data lives on your hardware, in your building, and there is nothing of yours on our side to delete. Parting ways means we stop showing up. It cannot mean losing access to your own information, because the architecture never gave us custody of it in the first place.
Is this as good as ChatGPT?
For the deepest frontier reasoning — no, and we won't pretend otherwise. The frontier labs' newest models are better at the hardest tasks. What your stack does is route honestly: local models handle the private and the routine, which is most of a working day, and when a task genuinely needs the frontier, you can still call it — on your terms, under your data rules, as a tool you choose rather than a landlord you rent from. You lose a little peak capability on rare tasks. You gain ownership of everything else, permanently.
Who is behind this?
Jeremy Ward. The system described on this page is the one he runs daily — a working institution with a pilot tenant and proving ground where every mechanism above earned its Live pill before appearing here. This is a personal practice with a public record, not a staffed agency wearing a founder's name. That's also why seats are capped: the promise never exceeds the person.
See It Running.
Not in a Deck.
Thirty minutes. One question — build it, or have it built? If it's not a fit, we'll say so in the first ten.
contact@riot-os.comNo pitch. No deck. Just a real conversation.
Want proof before people? Ask for the live 30-minute walkthrough — a screen-share of sanitized dashboards, the running system measuring itself in front of you. Nothing staged, nothing tunneled public.
Not ready to talk? Read the manifesto — and sign it if it says what you've been thinking. Name and email, no drip sequence, no gated anything. It's the only list we keep.