WSO2 Agent Manager reaches GA, an open-source control plane for AI agents across frameworks
Image: InfoQ
Why it mattersA team running agents in more than one framework, or planning to move between frameworks, can put the identity, permission and audit layer in one place instead of rebuilding it inside each stack.
WSO2 released its Agent Manager into general availability in September 2026, InfoQ reported on 18 September, six months after the platform entered beta in June. The repository at github.com/wso2/agent-manager carries an Apache-2.0 licence and describes itself as "an open control plane designed for enterprises to deploy, manage, and govern AI agents at scale."
What the control plane does
InfoQ describes Agent Manager as providing "centralized governance, identity management, security controls, and operational oversight for AI agents running across different models, frameworks, and deployment environments." Each agent gets a verifiable identity, role-based access, and delegation and token exchange between agents. The lifecycle side moves an agent through development, staging and production stages, and a Kubernetes-native sandbox runs the agent code with those controls attached.
WSO2 says the platform ships more than 40 built-in runtime controls, including personally identifiable information masking and rate limiting, and combines rule-based checks with LLM-based evaluation for monitoring agent behaviour. Traces flow through OpenTelemetry, so an existing observability stack sees the agent calls the same way it sees any other service.
What it interoperates with
Named integrations at GA are LangChain, CrewAI, Amazon Bedrock, Azure and Ballerina, plus support for custom-built agents. Standards support includes Model Context Protocol for tool calls and OAuth 2 extensions for the delegation and token-exchange flows. The framework and cloud list is the point of the release, because the argument WSO2 is making is that governance should sit outside the framework the agent runs in, so a team that swaps LangChain for CrewAI in six months does not have to rebuild identity and access on the way.
Where this fits against the platform vendors
InfoQ frames the gap this way: cloud vendors including AWS and Microsoft add governance features to their own agent platforms, but "no single platform yet provides a complete governance stack" that works across models, frameworks and clouds. WSO2's differentiation, in InfoQ's words, is that it "separates agent governance from agent logic," which is what makes the same set of policies apply whether the underlying agent is an Amazon Bedrock flow, a LangChain script or a Ballerina program.
A team already committed to a single agent framework and cloud may find the platform's own governance sufficient, and the trade against Agent Manager is one more service to run. A team with agents in more than one framework, or one that expects to move, has the opposite calculation: keeping identity, auditing and rate limiting inside each framework means writing them each twice, and each rewrite is where a policy drifts.
Source
Reported by: InfoQ
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
