Managed deployment — available now
We scope it, install it on your hardware, configure it for your workflows, and maintain it within an agreed support window. This is what this page sells.
CastelOS, installed on your machine and kept running.
The system itself — a local AI workstation set up for your workflows, installed and maintained by the person who built it. Not a cloud service: it runs on hardware you control, and it keeps running if your internet doesn't.
local deployment · operator-installed · from 3 900€
AVAILABLE NOW
What you can get today
Self-serve edition — in development
A version you install and run yourself, without us. It isn't available, and there's no date. If that's what you want, join the waitlist rather than buying this. →
WHAT YOU'RE BUYING
What you're actually buying — and from whom
Castel Studio is one person. For a system you'd depend on, that's the first thing you should know, not the last.
1 · Documented, not locked in
The setup is written down: what's installed, where the credentials live, which models and licences are in play, and how to export your data. If we part ways, you can hand the document to anyone.
2 · Support windows, not 24/7
Maintenance happens within a support window we agree in writing before deployment. We don't sell an enterprise SLA and we won't pretend to — if you need guaranteed round-the-clock cover, this isn't the right fit.
3 · Your hardware, your data
It runs on machines you own. You hold the data; you decide what goes into the system. We hold admin access for maintenance, and the scope of that access is in the agreement.
WHO THIS IS FOR
Who this is for
Teams that want the whole system rather than one service — and can live with a local deployment maintained by a named person.
You want it all in one place
Documents, automation, media — not four vendors and four bills.
Cloud is not an option
For reasons you don't need to justify to us.
Nobody in-house to run it
You want the system, not a new hire to babysit it.
You want a name, not a queue
The person who installs it is the person who answers.
WHAT IS INSTALLED
What is installed
- CastelOS installed on your hardware, configured for the workflows we agree
- A written deployment document: what runs where, credentials, model and licence profile, export paths
- Training for the people who'll use it
- Maintenance within an agreed support window
- An exit path — how to keep your data and the system if you stop working with us
How we work
- 01Scoping call (20 min)
- 02Written scope + fixed quote
- 03Deployment on your hardware
- 04Acceptance test + training
- 05Operation within the agreed window
PILOT OPERATING MODEL
How the pilot is judged, operated and handed over
A CastelOS pilot is not a loose bundle of AI tools. It is a small operating system with defined studios, human ownership, acceptance criteria and an exit package from day one.
PILOT ARCHITECTURE
The first deployable shape
- IntakeOne agreed entry point where tasks, documents and requests arrive.
- StudiosConfigured capabilities for knowledge, drafting, automation, media or analytics, only where in scope.
- ControlApproval gates, logs and support boundaries so the system does not act invisibly.
- HandoverA deployment document describing what runs where and how to keep or exit the system.
HUMAN ROLES
Who owns what
- Client owner
- Decides what success means and approves risky outputs.
- Operator
- Uses the system day to day and reports friction.
- Castel Studio
- Installs, documents, trains and maintains within the agreed window.
SUCCESS CRITERIA
What makes the pilot worth keeping
- The first workflows run with fewer manual handoffs.
- Users can find documents, draft outputs or route tasks without waiting for a specialist.
- Risky actions stop at a human checkpoint.
- The client can read the deployment document and understand the exit path.
EXIT PACKAGE
What remains after the pilot
Deployment note: machines, services, credentials location, models and licence profile.
Operations note: daily use, restart steps, support window and known limits.
Decision note: keep, expand, pause or hand the system to another engineer.
Illustrative pilot model. The actual success criteria and exit package are written into the scope before deployment, not improvised at the end.
ORIENTATION
What a pilot roughly costs
Pick the shape of your deployment and get an honest range — not a quote, an orientation. The number that matters comes out of a scoping call, in writing.
orientation range
3 900 – 5 400 €
deployment fee — hardware and monthly operation are separate
Request a scoped proposalWhat the range assumes
- Your hardware — if we source it, it's quoted transparently and separately, without a markup.
- Install, configuration for the agreed workflows, training and a written deployment document are included.
- A business-hours support window; wider cover is quoted separately.
- Monthly operation, if you want us to run it, is agreed separately.
- A written acceptance test before deployment — if it isn't met, the deployment fee isn't due.
This is an orientation range based on the assumptions above, not an offer or a quote. Every binding figure comes from a written proposal after a scoping call.
PRICE & ACCEPTANCE TEST
Book a scoping call
from 3 900€ · broader deployments quoted after scoping
Founding-client rates — reduced while we build the public portfolio. Limited slots.
From 3 900€
a narrow deployment on a single workstation: your hardware, agreed workflows, install, training and handover.
Broader deployments are quoted after the scoping call
what moves the number: how many workflows, how many machines, whether hardware is yours or we source it, how sensitive the data is, and how wide a support window you need.
Hardware, if we source it, is quoted transparently and separately — we don't mark it up.
The acceptance test
We agree an acceptance test in writing before deployment, with a criterion you can check yourself. If the deployment doesn't meet it, you don't pay the deployment fee.
FAQ
Common questions, honest answers
Is this a cloud service?
No. It's installed on hardware you control and runs locally. There's no tenant, no shared instance, and nothing of yours sits on our machines.
What happens if you're unavailable?
Everything is documented — install, credentials, licences, export paths — so another engineer can pick it up without us. We're one person: we don't sell 24/7 cover, and we say so before you buy rather than after.
Who owns what?
The hardware and the data are yours. You get the right to run the deployed system as agreed in the contract, and the deployment document is yours to keep.
Are the models licensed for commercial use?
That's confirmed before deployment, not after. The licence profile of what we install is written into the deployment document, and we don't install anything that isn't cleared for your use.
What if we want to leave?
You keep your data and the deployment document. The exit path is agreed at the start, not negotiated when relations are strained.
Why no fixed price?
There is one — from 3 900€ for a narrow deployment. Beyond that, the honest answer is that a pilot's cost depends on your hardware and scope, and a number invented before the call would be a worse promise than no number.
What this is not
- Not a cloud subscription — it's a machine in your building.
- Not an enterprise SLA — support windows are agreed, and there's no 24/7 cover.
- Not fully autonomous — it runs the workflows we configure, and a person stays in charge of them.
- Not a replacement for your team.
- Not self-serve — that edition is in development, without a date.
Book a scoping call
Book a scoping call
Prefer email? contact@castel.studio
Support windows are agreed in writing before deployment. There's no guaranteed round-the-clock cover, and we say so before you buy.