An industrial AI laboratory building the real-time decision layer for heavy industry — decentralized AI agents that coordinate people, machines and data in under a second.
Monolithic ERP reacts in batches. Our agentic AI contours react in real time — each agent owns its slice of the operation, negotiates with its neighbours and keeps working when the network doesn't.
Event-driven agent mesh replaces batch scheduling. Decisions propagate through the contour in under a second.
No single point of failure. Agents degrade gracefully and re-synchronize when connectivity returns — built for field and plant conditions.
Runs on-premise, at the edge or in the cloud. LLM-backbone independent: agents keep their contracts whatever model sits underneath.
Every contour is built on MAIF — our open multi-agent framework — so integrations are auditable and vendor lock-in stays out.
We state statuses plainly: what is live, what is a prototype, what is a proposed pilot.
An open standard for building auditable industrial agent systems: agent contracts, communication protocol, orchestration semantics. Registered and released under Apache 2.0.
github.com/ciriycpro/MAIFFirst-edition draft of a national preliminary standard (ПНСТ) for industrial multi-agent systems — developed within technical committee TC 164 "Artificial Intelligence" and open for public review.
github.com/ciriycpro/pnst-industrial-masWe treat multi-agent systems as the practical route to general artificial intelligence in the physical economy — superintelligence that arrives grounded, auditable and standardized, not as a black box.
