WSO2 releases open-source control plane for managing AI agents across enterprise environments

Agent Manager, launching in beta under an Apache 2.0 license, gives organisations a single point of visibility and governance over autonomous agents running across different frameworks, clouds and on-premises systems

Rania Khalaf, Chief AI Officer at WSO2

WSO2, an enterprise software company that builds middleware for identity management, API management and integration, has announced the beta launch of WSO2 Agent Manager, a platform designed to give organisations centralised control over AI agents operating across their systems. The product is scheduled for general availability in June 2026 and will be released under the Apache 2.0 open-source license.

The platform addresses a specific operational problem. As companies move from AI experimentation to production deployment, the number of autonomous agents running across an organisation grows, often built on different frameworks, deployed across different environments and managed by different teams. Without a centralised layer to track which agents exist, what they can access and how they behave, organisations face what WSO2 describes as "agent sprawl," where agents operate with limited coordination, inconsistent security controls and growing compliance exposure. According to Gartner, more than 40% of agentic AI projects are expected to be cancelled by 2027 due to rising costs, unclear value and insufficient risk controls.

Agent Manager functions as a single control plane that sits above the individual agent frameworks. It allows organisations to register and manage agents regardless of where they run (cloud, on-premises or hybrid) or which framework they were built with, including LangGraph, CrewAI and Ballerina. The platform assigns identity to each agent, controls what data and systems that agent can access, enforces governance policies across agents and the language models they use, and provides tracing and monitoring tools to track agent behaviour over time. Agents running outside the platform can also be brought under the same governance layer through federation.

On the technical side, the platform is built on Kubernetes with a zero-trust architecture, meaning each agent runs in isolation with its own access controls and lifecycle management, including the ability to intervene in real time if an agent behaves unexpectedly. The system is built on open standards including OpenTelemetry for observability, OpenAPI for interface definitions and Model Context Protocol (MCP) for agent-to-tool communication.

"AI agents introduce a fundamentally new challenge. Their autonomy and probabilistic behavior make them powerful but also difficult to control," said Rania Khalaf, Chief AI Officer at WSO2. "With WSO2 Agent Manager, we're bringing agents into the enterprise fabric where they are no longer invisible processes, but identified, governed, accountable entities that can be securely operated at scale."

The open-source licensing model means organisations retain full ownership of their technology stack and are not dependent on WSO2 to continue operating agents managed through the platform. WSO2 will showcase the product at WSO2Con North America, running May 20 to 22 in Austin, Texas.