Agents are ultimately decision makers. They make those decisions by combining intelligence with context, ultimately meaning they are only ever as useful as the context they can access. An agent that can't check inventory levels, look up customer history, or verify pricing isn't autonomous — it's just a chatbot that sounds confident. The real value of agentic AI comes from connecting reasoning capabilities to the operational data and actions that drive your business.

But getting context to agents is harder than it sounds. Your enterprise data lives behind APIs, event streams, databases, and services — each with different schemas, authentication requirements, and access patterns. Someone needs to understand each data source, define how it should be exposed to agents, handle credentials, manage errors, and figure out how to host the resulting integration. Multiply that by the dozens of systems that hold the context agents need, and you've got a project that never ships — even as pressure to deliver agentic capabilities accelerates.

Meanwhile, Kong already knows about much of this infrastructure. It knows the endpoints, the schemas, the authentication requirements, and critically, it knows who has access to what. That context has been sitting there, waiting to be activated.

With Context Mesh, it’s activated.

Context Mesh solves the agent context problem by leveraging what Kong already knows about your infrastructure. With the tech preview release, we're starting with how MCP interacts with APIs — the most common abstraction layer for enterprise data — and expanding from there. Here's what you can do:

• Discover what you have. Context Mesh automatically surfaces every API currently under Kong management. No manual inventory. No spreadsheet archaeology. You see your complete API landscape in one place, with full visibility into endpoints, schemas, and access patterns.
• Understand who can access what. Because Kong already manages authentication and authorization for your APIs, Context Mesh knows exactly which data sources any given user or service has access to. This becomes the foundation for agent-appropriate access controls — you're not starting from scratch on permissions.
• Curate your agent toolkit. Not every data source belongs in every agent's toolbox. Context Mesh lets you filter across endpoints from different APIs and select precisely which context you want to expose. You're composing a purpose-built toolkit, not dumping your entire data landscape into an agent's context window.
• Generate agent-ready tooling automatically. Once you've selected your data sources, Context Mesh generates fully-functional tool definitions that agents can consume. The schemas are correct. The authentication is handled. The error responses are meaningful. What would take days of manual work happens in minutes. We're starting with MCP as the output format — it's where the ecosystem is converging — but the abstraction is designed to support however agents consume context as the landscape evolves.
• Deploy to a Kong Gateway. Generated tooling isn't an artifact you have to figure out how to host. Context Mesh deploys directly to any of your Kong AI Gateway environments to host your agent tooling and enforces policies at execution time. You get production-grade infrastructure from day one.
• Define policies and orchestration logic. Define conditional policies and logic that govern how these tools behave. Rate limiting, conditional routing, data transformation — the same governance patterns you've built for APIs now apply to your agent infrastructure, enforced by Context Runner at runtime.

Individual pieces of this vision exist elsewhere. You can find tool generators. You can find API catalogs. You can find identity management solutions. What you can't find is a platform where these capabilities compound.

• Registry integration. Every MCP toolkit and server you generate through Context Mesh can be cataloged in Kong's registry. This isn't a separate system to maintain; it's automatic. Your agent developers discover available tools through the same portal they use for APIs.
• Inherited access controls. Because Kong already knows which data sources any consumer should have access to, it automatically knows which agent tools that consumer can use. You don't rebuild your access model for agents; you extend the one you already have.
• Portal-driven access requests. Consumers can request access to agent tooling through the same developer portal and registration flow they use today. The experience is familiar. The governance is consistent. The audit trail is complete.
• Identity as a foundation. Kong Identity becomes your agent identity management solution. When you're defining policies and access controls in Context Mesh, you're referencing the same identity infrastructure that governs your entire connectivity layer. Agents authenticate the same way services do. Policies apply consistently across human and machine consumers.

Context Mesh and Context Runner both enter tech preview this quarter. If you're a Kong customer exploring agentic AI, talk to your account team about early access. If you're evaluating how to connect your existing data infrastructure to the agents that will increasingly drive your business, this is the conversation to have.

The transition from API-first to agent-first doesn't require abandoning what you've built. It requires a platform that can bridge the two — discovering the context you have, transforming it into what agents need, and governing access with enterprise-grade controls.

That's what Kong Context Mesh delivers. And we're just getting started.