Ask most developers how their AI agent authenticates to external services and the answer usually involves a config file, an environment variable, or a token that shouldn’t be in version control but probably is. This works well enough in development. In production it’s one of the most common ways agent deployments get compromised.
AI agents need unusually broad access to be useful. They talk to APIs, read files, execute commands, connect to external services. That breadth makes them high-value targets. And unlike a traditional application with a single database connection, an agent might be holding credentials across dozens of integrations, many of them long-lived and unrotated.
OpenShift’s answer is built on open standards. The Zero Trust Workload Identity Manager brings SPIFFE and SPIRE to the platform, giving every agent workload a cryptographic identity without a single hardcoded key. The SPIFFE CSI Driver delivers those identities directly into workloads at runtime. The External Secrets Operator keeps Kubernetes secrets synchronized with an external vault, so credentials never need to live in the cluster directly. ServiceAccount token projection means every agent gets a short-lived token scoped to that specific workload, rotated automatically by the platform. Even at the UI level, OpenShift obscures secret values by default. Each of these tools is open source and composable, and together they give production agent deployments the identity story they actually need.
