Red Hat AI provides a platform-level solution for deploying AI agents, handling model serving, safety guardrails, agent identity, and persistent state. Using OpenClaw (an open source personal AI assistant) as a reference deployment, the post covers three inference paths: vLLM for self-hosted model serving via KServe, Llama Stack for unified multi-backend inference routing, and Models-as-a-Service (MaaS) for managed serving with built-in API key management and rate limiting. Agent identity is addressed through Kagenti, which introduces AgentRuntime CRDs for lifecycle visibility and AuthBridge for zero-trust service-to-service authentication using SPIFFE/SPIRE and short-lived JWTs. OpenShift enforces security defaults including non-root containers, built-in OAuth, and automatic TLS. A community installer (openclaw-installer) automates the full deployment in about two minutes.
Table of contents
Model connectivity: Three paths to inferenceAgent identity and zero trustPlatform security: What OpenShift enforces by defaultDeploy OpenClawTake the next step with OpenClaw and Red Hat AISort: