Your company, agent-native. Ready for a guided walkthrough.
Foundry is the same stack we run internally: agent-native operations, on infrastructure you control, with a human board that keeps final decisions in human hands.
What Foundry actually is
Six things we can say about it without stretching, because each one is how we already run.
We run on it ourselves, daily
Artifocial has operated agent-native on this stack every day since June 2026 — the company that builds Foundry is run by it. Our build-in-public record is the receipt.
The org we actually operate
Foundry is the packaged version of the organisation we run, not a reference architecture drawn on a whiteboard. It improves weekly, because we are the ones living with the rough edges.
One machine you control
Designed to run on a single machine you control. Your company memory, your issue history and your agents stay on hardware you own.
Your board keeps the final say
Approval gates, spend limits and kill switches are how our own governance works, so they are how Foundry ships. Agents propose; humans decide what leaves the building.
Bring your own model credentials
You supply your own model keys, or run local models. We never resell inference, so there is no margin between you and the model you chose.
Memory the agents actually consult
A tiered company memory your agents read before they act — indexed, retrievable, and telemetered so you can see when it was used and when it was not.
What it is not
We would rather lose you here than after you have deployed it. This is the part every other pitch in this category leaves out.
It is not a silver bullet
Foundry does not make a business work that does not already work. It gives structure to work you can already describe. If a process is undefined, an agent will run it badly and faster.
It improves with data
Quality tracks what you feed it. An org with written-down process, real history and honest metrics gets compounding value; an empty vault gets a competent stranger who has read nothing.
It is not one-click, and we will not say it is
Today this is a compose file, a CLI and a set of templates you configure. A wizard is on the roadmap, not in the box. We would rather you arrive with accurate expectations.
It does not replace your people
The stack is built around a human board that keeps approval authority. Every deployment we run, including our own, has a person deciding what ships.
How you would start
The diagnostic scan is free. Paid tiers are quoted after we scope how your company actually works.
Start with the scan
Free
Fourteen questions in four blocks. You get a plain read of where work leaks out of your business and three moves you can make on your own — no tools, no help from us needed.
Take the scanSelf-serve bundle
Contact for tailored pricing
Your scan answers configure a Foundry bundle — compose file, seeded company vault, and the skills and adapters that match how you actually work. You deploy it on your own machine.
Full audit + board session
Contact for tailored pricing
We run the onboarding with you: map your real workflows, choose the councils and skills, then walk your team through a live board session. We will quote a ramp time once we have run enough of these to have a real number.
The MVP is ready to explore
Foundry v0.1.3 supports Apple silicon and Ubuntu 22.04.5 LTS x86_64. The latest clean release-artifact run passed all 18 scripted acceptance checks on Ubuntu 22.04.5 LTS x86_64. It is an MVP, not a one-click wizard, and it is ready for working-build walkthroughs.
Tell us how you would want to use it. We will reply with a working-build walkthrough or the next onboarding slot that fits your team.
No device identifiers, raw logs, or diagnostics are published on this page. The release proof is limited to a sanitized boundary: clean release-artifact install, then 18 scripted checks.