EONYX Agent is a terminal AI agent for companies and industrial enterprises. It works with production data — documentation, machines and ERP systems — so that engineers solve tasks faster without writing a single line of SQL or code.
This blog is the project documentation: installing and configuring the agent, memory, skills, MCP integrations, permissions and real-world scenarios inside a closed production network.
1 What the agent does
- Technical documentation — generates and updates maintenance cards, standard operating procedures and work instructions to ISO and IATF 16949 standards — based on your enterprise’s data, not on generic templates
- Diagnostics and predictive maintenance — analyzes logs from CNC machines, robots and production lines: finds anomalies, predicts failures and proposes ready-made PLC patches before the line stops
- ERP integration — connects to SAP, 1C and internal databases. An engineer asks a question in plain language — the agent runs the query, builds a report and, if needed, creates an order
- Enterprise context — the agent understands the structure of the production: areas, equipment, suppliers and standards. Answers are tied to specific assets, not to abstract knowledge
In short
One agent for the whole enterprise: the engineer writes the task as text — the agent reads the data, makes decisions and reports the result.
2 How it works
- Connection — ERP, machines, documentation
- Setup — the agent learns the structure of your enterprise
- Deployment — in your network, without the cloud
- Result — engineers do engineering instead of routine
The agent is deployed on a production server or on an engineer’s workstation — Windows and Linux, including fully offline. EONYX Agent is used by the production teams of Changan, Avtotor, Jetour, Zhejiang Sunshine Group and Haval; the product is available worldwide, including the CIS, MENA and Southeast Asia.
3 Principles
On-premise by default
The agent runs locally or inside a closed network. Enterprise data does not leave your network: no external cloud — full control over security and compliance.
A terminal instead of a complex UI
No bulky interface. The engineer writes the task in the terminal — the agent acts: reads the data, runs the queries and returns the result.
Control and permissions
Every action of the agent stays within explicit permissions. What is not allowed is not executed; anything dangerous and irreversible requires confirmation from the engineer.
Transparency
The configuration lives in the repository, not in someone’s head:
eonyx.json, AGENTS.md, skills in SKILL.md — all under version control and available to the team.