Eclipse LMOS & AGC : Agentic Computing Born from the Messy Reality of Enterprise Systems

Modern enterprises don’t live in clean labs , they run on decades of systems, workflows, messy APIs, and huge operational investments. When we tried to build AI agents inside that world in 2023, we found no real architecture or tooling for it. That’s how Eclipse LMOS and AGC (Agentic Compute) were born , from the messy, real-world constraints of a large European enterprise. We first defined agentic constructs on top of existing JVM systems, enabling engineers to build reliable agents within legacy environments. This evolved into LMOS Arc, a Kotlin-based framework, and LMOS ADL, a declarative layer that lets business teams iterate on agent behavior for reliability and control. As agent deployments scaled, we built AGC, a runtime for ephemeral and long-running agents, observable, debuggable, and compliant. Together, these open-source primitives power reliable enterprise agentic infrastructure, enabling brownfield teams to build on what they already have.