*Introducing Kong AI Registry: Governed discovery for enterprise AI models, MCP servers, and agents.*
AI agents are becoming increasingly capable, but their usefulness depends on what they can access. An agent that can reason through a customer support issue is of limited value if it can’t retrieve customer information, check an order status, or initiate a refund. Similarly, an IT operations agent needs access to monitoring systems, infrastructure APIs, and diagnostic tools if it’s going to do more than recommend what a human should investigate.
Giving agents access to these capabilities is becoming one of the central challenges of enterprise AI. Organizations have spent years building APIs, integrating applications, and making enterprise data accessible to developers. Now they need to make those same capabilities available to AI agents, often across different teams, environments, and technology stacks.
But there’s an important difference between making a capability available and deciding who, or what, should be allowed to use it.
With the **Kong AI Registry**, we provide a centralized way to register, manage, and govern the AI models, MCP servers, and agents available across the enterprise. It helps organizations move beyond manually configured integrations toward a model where developers and agents can discover sanctioned capabilities, while the enterprise maintains control over what’s available and how those capabilities are accessed.
Most AI agents today operate with a predefined set of connections. Developers configure which models, MCP servers, tools, and other agents they can interact with, often embedding those connections directly into the agent’s configuration or code.
That approach works when an agent performs a narrow set of tasks using a handful of known services. But as enterprises deploy more sophisticated agents, the limitations become apparent. An agent investigating a customer issue might need to retrieve account information, query a billing system, and invoke another agent to resolve a technical problem. The capabilities it needs may vary depending on the request, and new tools and services may become available over time.
Requiring developers to anticipate and configure every possible connection creates brittle architectures that are difficult to maintain. Adding a new capability can mean updating multiple agents, while changes to existing services can break established workflows. As the number of agents, MCP servers, and models grows, managing these connections individually becomes increasingly impractical.
Dynamic discovery offers a better approach. Rather than relying exclusively on predefined connections, agents can discover available capabilities and identify the resources appropriate for the task at hand. This allows organizations to introduce new models, tools, and agents without requiring every existing agent to be individually reconfigured.
But dynamic discovery introduces a critical security challenge. Just because an enterprise has an MCP server that can access sensitive financial records doesn’t mean every agent should be able to discover or invoke it. Different agents operate on behalf of different users, teams, and applications, each with their own permissions and responsibilities.
Enterprises therefore need more than a way for agents to find capabilities. They need a way to govern which capabilities are available for discovery and ensure that agents can only access resources they’re authorized to use.
That’s the challenge Kong AI Registry is designed to address.
Kong AI Registry addresses this challenge by providing a centralized place to register and manage enterprise AI capabilities, including AI models, MCP servers, and agents.
Platform teams can organize the capabilities they want to make available, while developers and AI agents gain a more consistent way to discover and consume them. Instead of requiring every team to maintain its own collection of connections and integrations, organizations can establish a governed set of reusable AI resources.
Kong AI Registry supports three categories of AI capabilities:
Developers can discover approved resources through Kong Konnect, reducing the time spent finding and integrating existing capabilities. For AI agents, programmatic discovery creates opportunities to identify available resources dynamically rather than relying exclusively on hard-coded connections.
Imagine an IT operations agent investigating an application outage. It might need to retrieve monitoring data, examine recent infrastructure changes, and invoke a specialized diagnostic agent. With a registry-based approach, these capabilities can be made available for discovery, allowing the agent to identify appropriate resources without every connection having to be independently configured.
As new models, MCP servers, and agents become available, organizations can introduce them into the governed ecosystem without requiring every development team to recreate the integration.
The objective is to move from a model where developers configure every connection to one where the enterprise governs the capabilities available for discovery, and agents can identify the resources they’re permitted to use.
Making AI capabilities discoverable solves an important development challenge, but it introduces another question: How do you prevent agents from accessing resources they shouldn’t?
Consider a financial services organization with multiple AI agents. A customer service agent might need access to account information and transaction history, while a financial analysis agent requires access to portfolio data and investment research. Both agents operate within the same enterprise environment, but their permissions should be very different.
A registry that simply lists every available capability doesn’t solve this problem. Without appropriate controls, it could make sensitive resources easier to find without ensuring they’re accessed appropriately.
This is why governance is fundamental to Kong AI Registry.
Kong AI Registry works with Kong’s broader identity, access management, and runtime security capabilities to help organizations control which AI resources are available and how they’re consumed. Platform teams can manage the availability of approved capabilities, while identity and access policies determine which users, applications, and agents are authorized to use them.
Importantly, **registering a capability doesn’t mean granting unrestricted access to it.** Organizations need to distinguish between a resource that exists, one that’s available for discovery, and one that a particular consumer is authorized to use.
Kong AI Registry is part of the broader Kong AI Management platform, where discovery, identity, access control, and runtime governance work together.
Registry helps organizations manage which AI capabilities are available for discovery. **Kong AI Gateway governs their consumption at runtime**, providing authentication, authorization, policy enforcement, traffic management, and observability.
For example, a platform team might register an MCP server that exposes customer account information. A developer building a support agent can discover that capability through the registry. When the agent attempts to invoke the service, Kong AI Gateway can authenticate the request, evaluate applicable access policies, and enforce the controls governing that interaction.
Organizations can also apply rate limits, security guardrails, and other runtime protections while monitoring how AI resources are consumed. Kong Token Vault further strengthens this architecture by enabling authorized access to downstream services without requiring agents to hold the underlying credentials themselves.
Together, these capabilities allow enterprises to support dynamic AI architectures without treating discovery as unrestricted access. Agents can find the resources they need, while the organization retains control over whether and how those resources can be used.
When approved models, tools, or services are difficult to find or integrate, developers may turn to external alternatives or create their own connections. Over time, organizations can accumulate unmanaged AI integrations, making it harder to understand which systems are being used and whether appropriate security controls are in place.
This is one of the factors contributing to shadow AI.
Security policies are necessary, but they aren’t sufficient on their own. If the approved approach is significantly harder than using an unmanaged alternative, developers will continue to look for ways around the friction.
Kong AI Registry helps address this challenge by making sanctioned AI capabilities easier to discover and reuse. A developer looking for an MCP server that provides access to enterprise data shouldn’t have to search internal documentation, contact several platform teams, or build a new integration from scratch.
Combined with identity-based access controls and runtime enforcement through Kong AI Gateway, Registry helps organizations establish an accessible, governed alternative to fragmented, ad hoc AI integrations.
As enterprises move from individual AI applications to interconnected agentic systems, the number of interactions between models, agents, tools, APIs, and enterprise systems will grow substantially.
Managing those interactions through individually configured integrations will become increasingly difficult. Developers need a more efficient way to discover and reuse capabilities, while platform and security teams need a consistent approach to managing what’s available and enforcing access.
Kong AI Registry provides an important piece of that infrastructure. Together with Kong AI Gateway, Context Mesh, and the broader Kong AI Management platform, it helps organizations connect agents to enterprise capabilities while maintaining control over access, security, and runtime operations.
The goal isn’t to give every agent access to everything. It’s to make the right capabilities available to the right agents, with the controls enterprises need to operate AI confidently at scale.
**Learn more about Kong AI Registry and how Kong AI Management helps enterprises connect, control, and operate AI at scale.**
API management solved a similar problem for developers. Instead of asking a platform team every time they needed an API, developers could go to a Developer Portal, discover what was available to them, and start building — within the governance bound
Agentic AI governance is the structured management of the authority an organization hands to autonomous AI systems that plan and take action on its behalf. It covers what an AI agent is allowed to do, which systems it can touch, and how those permis
If you've used KAi before, you'll notice the difference immediately. The enhanced version has deeper access to your Konnect environment and new tools for debugging, analytics, and configuration management: Active tracing and debugging: KAi can now
Every January, the same resolutions show up: eat better, exercise more, finally learn that language, finally figure out that production use case for AI agents (OK, this one isn’t so typical unless you operate in our universe). But if you're respons
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
Today, we're open-sourcing Volcano SDK, a TypeScript SDK for building AI agents that combines LLM reasoning with real-world actions through MCP tools. Why Volcano SDK? One reason: because 9 lines of code are faster to write and easier to manage tha
