CNCF: platform engineering must treat AI agents as equal platform consumers alongside applications

Why are agents becoming the platform’s third equal consumer?

Lakmal Warusawithana, an engineer at WSO2, writes on the CNCF blog about how platform engineering — the discipline of building internal platforms that make it easier for development teams to ship software — must expand beyond its current focus on applications and infrastructure resources. According to him, AI agents are increasingly calling APIs, consuming resources, and triggering workflows on their own, so the platform must treat them as an equal, third consumer alongside applications and resources, rather than an add-on bolted on afterward.

OpenChoreo as an example of unified governance

As a practical illustration, Warusawithana cites OpenChoreo — an extension of the Internal Developer Platform (IDP), the internal platform that gives teams standardized access to tools for development, deployment, and monitoring. OpenChoreo offers human users and AI agents shared interfaces, such as an MCP server — a protocol that gives agents structured access to the platform’s tools and data — instead of separate, incompatible access points for each type of user.

One platform versus separate access paths

The traditional approach manages applications and resources through a unified platform, while agent access is handled afterward and ad hoc, often through special integrations outside the standard governance layer. Warusawithana advocates the opposite model: applications, resources, and agents are managed consistently within the same platform, with the same access rules, oversight, and accountability for all three categories of consumers.