Oracle recently introduced the Open Agent Specification (Agent Spec) which is a framework-agnostic, declarative standard designed to make AI agents and workflows portable, reusable, and executable across any compatible framework.
https://blogs.oracle.com/ai-and-datascience/post/introducing-open-agent-specification
A LangGraph adapter is already available, enabling initial interoperability - agent-spec/adapters/langgraphagentspecadapter at main · oracle/agent-spec · GitHub
However, it would be highly valuable to see native support within LangGraph for importing and exporting agents using this specification. Such integration could pave the way for a truly open Agent Registry, much like how ONNX standardized model exchange across deep learning frameworks.