Red Hat is betting on AgentOps to close the gap between AI experiments and production

Red Hat announced RHAI 3.4 at Red Hat Summit in Atlanta, adding a governed Model-as-a-Service layer for enterprise AI deployments.

The update also improves distributed inference and model serving performance, including support across vLLM and llm-d.

New AgentOps capabilities add tracing, observability, evaluations, and lifecycle management to help teams run agents in production.

RHAI 3.4 also includes MCP access and SPIFFE/SPIRE-based identity controls to strengthen governance and security in hybrid cloud environments.