Kong Agent Gateway Is Here — And It Completes the AI Data Path
When enterprises started experimenting with LLMs, [the governance challenge](https://konghq.com/solutions/ai-governance)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.
That world is over. Today's agentic architectures look nothing like that. Agents are calling tools via [MCP](https://konghq.com/blog/learning-center/what-is-mcp)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.
Kong Agent Gateway is a new capability within Kong AI Gateway that extends our platform to more robustly cover agent-to-agent (A2A) communication.
With this release, Kong AI Gateway now natively handles all three patterns of AI traffic in a product that is already being used by some of the largest enterprises in the world to govern real, production AI workloads. As of 3.14, Kong AI Gateway can be implemented as an:
Combined with Kong's API Management and Event Management capabilities, you now have a single governance layer across the entire stack — APIs, events, LLM calls, MCP tool access, and A2A communication. We call this the full AI data path, and no other vendor covers it.
Yes, A2A is still early. The protocol launched in April 2024, backed by Google, but enterprise adoption is now accelerating fast given the [pressures to start innovating with agentic AI](https://thenewstack.io/agentic-ai-connectivity-platform/)pressures to start innovating with agentic AI.
Organizations that are serious about moving agentic workloads into production are already hitting many of the same walls they hit when trying to hardwire apps to LLMs: how do you govern communication agent-to-agent communication between agents you didn't hand-wire together?
Without A2A governance, enterprises are stuck choosing between stitching together point solutions, building custom proxies, or accepting blind spots in their AI infrastructure. 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](https://konghq.com/blog/enterprise/agentic-ai-governance-managing-shadow-ai-risk)kill agentic AI velocity.
And as Gartner put it in the [Emerging Tech Adoption Radar 2026](https://www.gartner.com/document-reader/document/7322130?ref=solrAll&refval=534918185&)Emerging Tech Adoption Radar 2026:
*"Discoverability, negotiation and transaction between agent-to-agent (A2A) counterparts represent the next horizon — an emerging class of differentiated features for AI gateways."*
But Gartner goes even further, noting that AI gateways and agent management platforms together "form the backbone of safe and scalable AI adoption" and help organizations contain agent sprawl. Kong Agent Gateway is designed to be exactly that backbone for enterprise [AI governance](https://konghq.com/solutions/ai-governance)AI governance. And it’s now a core part of the larger Kong Konnect platform as an [agent management platform](https://konghq.com/blog/enterprise/agentic-ai-developer-platform)agent management platform, making it a leading choice when conducting an enterprise AI governance platform comparison.
Platform and infrastructure teams can define global policies that govern which agents can communicate with each other, what data can flow between them, and how every interaction is logged. Here's what that looks like in practice:
When we build Agentic AI infra, we test it on real agents. If you're looking for real-world agentic architectures examples, check out the video below that shows Kong Agent Gateway (along with the rest of the Kong platform, including Event Gateway) being used to govern real-time multi-agent communication as a part of a larger Agent, Orchestratrator, and subagent pattern.
Remember to think about *more *than just native AI traffic.
As massive as the Agent Gateway is, A2A communication is just *one part *of the larger AI data path. When planning out your multi-agent strategies, make sure that you have plans to govern the entirety of the intelligence (LLMs) and context (MCP, APIs, events, data, other agents) surface that your Agents will need to traverse. If you’re interested in how to combine Agent Gateway with other parts of the Kong platform, we recommend checking out the following resources:
Kong Agent Gateway is available now for all Kong Konnect customers as part of Kong AI Gateway. If you're already using Kong API Gateway and/or AI Gateway for API and LLM traffic, you already have a natural path to Agent Gateway as your agentic workloads mature.
Learn more and get started at [konghq.com/ai-gateway](https://konghq.com/ai-gateway)konghq.com/ai-gateway.
What is the A2A protocol?
The Agent-to-Agent (A2A) protocol is an emerging standard launched in April 2024 (backed by Google) that allows autonomous AI agents to discover, negotiate, and transact with one another. It provides a standardized framework for agents to share context and delegate tasks without requiring human-in-the-loop intervention or manual hardwiring by developers.
How does Kong Agent Gateway compare to OpenAI Gateway?
While tools like the OpenAI Gateway are excellent for managing linear traffic to specific LLM models, Kong Agent Gateway is designed to govern the *entire* AI data path. This means Kong doesn't just manage LLM traffic; it also secures agent-to-agent (A2A) communication, governs Model Context Protocol (MCP) server access, and integrates with your existing APIs and event streams—making it a more comprehensive enterprise AI governance platform.
How do I prevent prompt injection between AI agents?
To prevent prompt injection in multi-agent systems, you need a governance layer capable of real-time traffic inspection. An Agent Gateway sits between communicating agents to inspect message content in flight. It evaluates the data being passed via the A2A protocol and blocks malicious payloads, prompt injection attempts, or policy violations before they can propagate to other agents or LLMs.
How can I track token costs per AI agent?
Tracking costs in a multi-agent architecture requires an AI gateway that supports cost allocation by agent identity. Kong Agent Gateway authenticates every agent participating in a workflow and tracks its specific token consumption and resource usage. This allows platform teams to allocate costs accurately per AI agent, rather than just seeing a massive, undifferentiated bill from an LLM provider.
What are the best AI governance tools for multi-agent systems?
The best AI governance tools for multi-agent systems must handle more than just LLM rate limiting. A complete solution requires an LLM Gateway for model traffic, an MCP Gateway for tool and data source access, and an Agent Gateway to secure the A2A protocol. Platforms like Kong Konnect combine these elements to provide unified observability, identity enforcement, and audit logging across the full AI data path.
The Shifting Economic Landscape: The AI token economy in 2026 is evolving, and enterprise leaders must distinguish between low-cost input tokens and high-premium output tokens to maintain profitability. Agentic AI Financial Risks: The transition t
[](https://konghq.com/blog/enterprise/ai-input-vs-output-cost-management)
Managed Redis cache is a turnkey "Shared State" add-on for Kong Dedicated Cloud Gateways. It is designed to combine the performance of an in-memory data store with the simplicity of a SaaS product. When you spin up a Dedicated Cloud Gateway in Kong
[](https://konghq.com/blog/product-releases/multicloud-cloud-gateways-managed-redis-cache)
The Kong MCP Registry acts as a central directory for AI agents and clients to access services that provide context or take action. For AI agents, think of it as a combination of a "Service Catalog" and a "Developer Portal." It offers the metadata,
[](https://konghq.com/blog/product-releases/kong-mcp-registry-tech-preview)
MCP ACLs, Claude Code Support, and New Guardrails New providers, smarter routing, stronger guardrails — because AI infrastructure should be as robust as APIs We know that successful AI connectivity programs often start with an intense focus on how
[](https://konghq.com/blog/product-releases/ai-gateway-3-13)
Bring Financial Accountability to Enterprise LLM Usage with Konnect Metering and Billing Showback and chargeback are not the same thing. Most organizations conflate these two concepts, and that conflation delays action. Understanding the LLM showb
[](https://konghq.com/blog/enterprise/llm-cost-management-ai-showback-and-chargeback)
A Response to Gartner’s Latest Research We have crossed a threshold in the AI economy where the competitive advantage is no longer about access to data — it’s about access to context. The "context economy" has arrived, defined by a fundamental
[](https://konghq.com/blog/enterprise/the-platform-enterprises-need-to-compete)
What does the solution space look like so far? The solution landscape is complicated by the fact that MCP is still finding its footing, and there are many various OSS projects and vendors that are rapidly shipping “MCP support” in an attempt to take
[](https://konghq.com/blog/product-releases/enterprise-mcp-gateway)