A new open Agent Definition Language is transforming enterprise AI by simplifying how intelligent agents are designed, deployed, and governed—enabling faster, more transparent, and scalable agentic systems built on open standards.
The Eclipse Foundation has introduced Agent Definition Language (ADL) to its Eclipse LMOS (Language Models Operating System) project, marking a major milestone in open-source Agentic AI. Positioned as an industry-first, ADL offers a structured, model-agnostic framework that allows enterprises to design, deploy, and scale intelligent AI agents collaboratively and transparently.
Designed to address the complexity of prompt engineering, ADL lets business and technical teams co-define agent behavior in a consistent, versionable way. This approach enhances reliability, governance, and scalability across complex enterprise environments. Unlike proprietary platforms, Eclipse LMOS provides a vendor-neutral foundation built on open standards such as Kubernetes, enabling interoperability and operational flexibility.
The project comprises three key components:
- Eclipse LMOS ADL: A structured language and toolkit enabling domain experts to define and visualize agent behavior.
- Eclipse LMOS ARC Agent Framework: A JVM-native, Kotlin-based runtime with tools for developing and debugging AI agents.
- Eclipse LMOS Platform: A CNCF-based orchestration layer managing the lifecycle, routing, and observability of multi-agent systems.
ADL’s introduction strengthens LMOS’s role as an enterprise-grade platform already powering large-scale deployments such as Deutsche Telekom’s Frag Magenta OneBOT, which handles millions of customer interactions across Europe.Industry analysts see this as timely. According to Gartner, by 2028, 15% of daily business decisions will be autonomously made through agentic AI, while a third of enterprise apps will embed such capabilities. LMOS bridges this shift by enabling IT teams to leverage existing DevOps tools like Kubernetes and Istio, accelerating AI adoption without disrupting infrastructure.
By allowing non-technical users to encode domain logic directly into agents, ADL makes AI system design as intuitive as process modeling. “We wanted to make defining agent behavior as natural as describing a business workflow,” said Arun Joseph, LMOS project lead.Open, modular, and cloud-native, Eclipse LMOS positions itself as a practical, extensible path for enterprises to build scalable, collaborative, and transparent multi-agent systems—ushering in a new era of open Agentic AI.
