Artekis Security: Read-Only AI Analysis for Ignition

What it never does

It cannot act on your process

It never writes to a device, OPC item, setpoint, expression tag or alarm.
It never acknowledges or shelves an alarm.
It never applies a configuration change. A proposed change is a recommendation your team applies under its own change control.
It never creates or alters database tables. Your DBA applies its database script.
It has no cloud component, it is not an MCP server, and it trains nothing.

What leaves the gateway

The model key lives in Ignition’s Secret Provider and is referenced by name. Artekis never stores it or shows it.

Anthropic, or any OpenAI-compatible endpoint, including a model on your own network so nothing leaves it.

Each model call is priced and listed, with Cost and Egress pages showing spend and every message sent.

Show prompt displays exactly what the model would receive for a run, free, with nothing leaving the gateway.

Who can do what

Security levels decide who can read findings, ask questions, edit context models and upload documents, and the gateway checks again on every call.
Ask is for signed-in users only, with per-user limits on questions an hour and spend a day.
Spend caps per provider per day and per run; a run that would exceed one is skipped and says so.
An off switch per plant area stops every run on it before anything is gathered or sent.
Documents are cleaned before storage: key-shaped strings and long encoded runs are redacted, and every upload is audited.
Your data stays in your database: runs, findings, evidence and verdicts, in your SQL Server, MariaDB, MySQL or PostgreSQL.

Next step

Send us your security questions. We can go through the configuration, the data it sends and the audit trail.