Kong launches Agent Gateway for multi-agent AI traffic

Kong has launched Agent Gateway in AI Gateway 3.14, adding support for agent-to-agent AI traffic.

The release expands Kong's AI Gateway beyond large language model traffic and Model Context Protocol integrations. It targets organisations building systems in which software agents interact with tools, hand off tasks to other agents and generate event streams with limited human oversight.

Agent Gateway is intended to give engineering teams a single place to govern AI traffic across LLM, MCP and agent-to-agent communication. Kong said businesses moving AI systems from pilot projects into production face growing pressure to monitor usage, control costs and maintain audit trails across increasingly complex workflows.

Alex Drag, Head of Product Marketing at Kong, said the governance challenge has shifted as AI systems have evolved from simple model requests into multi-step agent interactions.

"When enterprises started experimenting with LLMs, the governance challenge, while still difficult, was relatively contained once you had an AI Gateway for LLM traffic. You had a request going to a model, a response coming back, and a gateway in between to enforce policy. With the right solutions, this becomes manageable pretty quickly," Drag said.

"That world is now over. Today's agentic architectures look nothing like that. Agents are calling tools via MCP. Agents are delegating tasks to other agents via A2A. These other agents are producing and consuming event streams. Data flows in every direction, often with little to no human in the loop, and often with no visibility into what's actually happening."

"This is the problem we built Kong Agent Gateway to solve. And now Agent Gateway is GA and prod-ready as of AI Gateway 3.14."

Broader control

The product extends governance across what Kong describes as the full AI data path, covering models, applications, tools and data exchanges. In practice, that means routing and observing traffic between AI agents, as well as the APIs and event-driven systems they use.

Kong tied the launch to a broader push by large businesses to operationalise AI. Rather than sending prompts to a single model, many are now orchestrating multiple services and software agents, creating a larger attack surface and making internal oversight more difficult.

Reza Shafii, senior vice president of product at Kong, said customers are encountering three recurring problems as they try to deploy these systems.

"Every enterprise is wrestling with the same three challenges: not having full visibility into all AI traffic and resource consumption within an agentic workflow, struggling to adopt AI in a way that helps increase margins, and dealing with issues when moving AI and agentic workloads into production," Shafii said.

"We built Agent Gateway to solve this directly. Engineering teams can now govern all of their multi-agent traffic in a single place. This is the kind of control and connectivity that makes agentic AI workable at enterprise scale."

Operational focus

The product includes a unified observability dashboard in Kong's Konnect platform, along with security controls, access management and audit functions aimed at production deployments. It also provides tracking of token consumption and resource usage across agent workflows, which companies can use for cost allocation and internal controls.

The emphasis on audit logging reflects a wider concern in corporate AI deployments: proving how a decision or action was produced. As software agents begin to invoke tools and call other agents, that chain can become harder to follow than a single user prompt to a model.

Kong argued that many rival tools cover only one part of that traffic, forcing customers to either combine separate products or rely on custom-built proxies. Those gaps, it said, can leave organisations without a full picture of data flows inside agent-based systems.

Drag said those blind spots create operational and security risks.

"Without A2A governance, enterprises are stuck choosing between stitching together point solutions, building custom proxies, or accepting blind spots in their AI infrastructure," Drag said.

"These blind spots in agentic architectures are particularly dangerous, as they can lead to rogue agents consuming massive token budgets or leaking sensitive enterprise data across unauthorized boundaries. None of those options scale, and none of them help you avoid the security pitfalls that will kill agentic AI velocity."

Market context

Kong also cited Gartner research on the role of AI gateways in policy enforcement and usage monitoring. It linked that view to the growing importance of agent-to-agent interactions, which are emerging as another layer of traffic that businesses may need to secure and observe alongside more established API and model connections.

The launch reflects a broader shift in the AI infrastructure market as suppliers move from model access tools toward systems for governance, monitoring and cost control. For buyers, the issue is becoming less about connecting to a single model and more about tracking complex chains of actions across agents, tools and services.

Agent Gateway is available as part of Kong AI Gateway in the Konnect platform.