Self-serve
Run it in your boundary.
One command installs the ArtzAIn decision engine and its dashboard on your own machine or VM. No account. No card. No vendor callouts — your data never leaves your boundary, and the evidence it seals verifies offline. You need Docker and Python 3.10+; the installer checks everything else and tells you the one-line fix for anything missing.
The command ends with your browser on the first-run page: set an email and password — that account is the platform admin — and land on the operator checklist. First sealed decision inside 15 minutes, without editing a single file. Both scripts are short and readable on purpose: install.sh · install.ps1.
Requirements
A laptop is enough.
What you get
The real engine, the real dashboard, your keys.
This is the same decision core the hosted platform runs — not a demo build. Every agent action is checked against policy before it runs and sealed into an Ed25519-signed, hash-chained audit ledger before the response returns. The operator dashboard, Roger, the policy catalog, the agent registry, and the full operator manual are served locally beside it. Signing keys are generated inside your boundary on first boot and never leave it; evidence exports verify offline with artzain audit verify — as SELF-ATTESTED until you hold a licence certificate, and that verdict never blocks.
The install runs the GPU-free core profile: local LLM chat, the coder fixer, and LightGBM propensity stay idle (they mark themselves degraded rather than pretending), while the decision path, sealing, policy enforcement, connectors, and evidence exports are all fully live. GPU inference in your boundary is an engagement conversation: hello@cognexuslabs.ai.
The trial, plainly
30 days. Then expansion pauses — governing never stops.
A fresh install runs a 30-day trial, counted from first boot and visible on the dashboard from day one. When it ends, connecting new sources, registering new agents, and minting new API keys pause until a licence certificate is imported. Everything already running keeps running: decisions keep sealing, existing connectors keep syncing, and your evidence keeps verifying — an engine whose job is governance does not hold your agents hostage over paperwork. Importing the certificate lifts the pause live, no restart.
Supply chain
Digests you can verify, not tags you have to trust.
The installer never pulls a floating tag. Releases publish to a public registry as multi-arch images whose digests are recorded in the stable-channel manifest and signed with cosign by the release workflow's own identity — verifiable by anyone:
Leaving is honest too: artzain local down stops the stack and keeps your data; down --purge destroys it only after you type the install's own id back, and reminds you to export your evidence first. Nothing phones home either way.