Every organization is in the process of building an AI and agentic strategy that will deliver meaningful impact to their business. While enthusiasm is at an all-time high, the percentage of organizations actually getting agents deployed into production is at an all-time low. A key contributing factor to this problem is that the cloud-native infrastructure we’ve used to support production applications over the last 10 years is simply not designed to address the new requirements of agentic applications. Meeting this challenge requires Kubernetes itself to be extended to become context-aware - treating agents, tools, and LLMs as first-class workloads and delivering the security, observability, and governance required for production-grade agentic infrastructure.
“As cloud native organizations embrace agentic AI, the real opportunity lies in moving beyond pilots to enterprise-wide impact. Kubernetes alone was not designed to handle the scale, complexity, and security demands of AI workloads,” said Paul Nicholson, Research VP, Cloud and Datacenter Networking at IDC.
We have been working with customers and the community to bridge this production gap and drive the future of context-aware networking with the donation of agentgateway to the Linux Foundation and context-aware runtime with kagent in the Cloud Native Computing Foundation. The next logical step to help organizations deliver production-ready agentic infrastructure was to bring these leading open source projects together into a fully supported, integrated platform complete with centralized management, observability, and governance. Today, we are excited to announce the availability of kagent enterprise - the first and only context-aware platform for agents, tools, and LLMs on Kubernetes. We’ll dive deeper into the details later in this blog, but the best way to get acquainted with kagent enterprise is to see it in action. Check out the demo video below:
Context awareness at every layer of agentic infrastructure
AI agents bring unique requirements around identity, cost transparency, observability, and governance. Until now, no single platform made it possible to build, debug, manage, and run agents across federated Kubernetes environments. Kagent enterprise allows teams to bridge the production gap by providing context awareness at every layer of an agentic infrastructure stack:
- Context-aware networking: kagent enterprise includes agentgateway, an agent-native data plane optimized for agentic AI connectivity with full support for MCP, A2A, and leading LLM provider protocols. Created by Solo.io and contributed to the Linux Foundation, agentgateway takes a more comprehensive approach to AI connectivity than any other AI gateway in the market, supporting LLM consumption, agent-to-agent, and agent-to-tool interactions across any tool server or agent framework.
- Context-aware runtime: kagent enterprise introduces a new runtime layer for agents and tools that extends Kubernetes to become context-aware. Unlike traditional cloud native runtimes that treat workloads as a black box, agentic runtimes require a new identity and policy model for agents acting on behalf of users, advanced failover and memory management for agents, and deeper observability instrumentation to explain and audit how agents and tools interact. Kagent enterprise has built-in support for creating and deploying agents and tools, but also integrates with alternative agentic frameworks (e.g. Agent Development Kit, Langchain) and any MCP-compliant tool server implementation. Launched in March 2025 and quickly adopted as a CNCF project, kagent has quickly grown to 800+ community members and 100+ contributors as the community-backed foundation for agentic infrastructure.
- Context-aware platform: kagent enterprise brings context-aware connectivity and runtime together with a centralized management plane to provide AgentOps - a unified platform for managing and securing agentic infrastructure. The AgentOps dashboard gives teams centralized visibility with an agent graph and end-to-end tracing of user, agent, tool, and LLM interactions. Policy and lifecycle management are built in, with declarative APIs and UI controls to create, deploy, update, and retire agents. An agent registry makes it easy to discover available agents and tools, while human-in-the-loop and human-on-the-loop controls provide the safeguards enterprises need to scale agentic applications with confidence.
Open source foundation for agentic infrastructure
At Solo.io, we believe the future of agentic infrastructure must be built in the open. Over the past year, we’ve worked with the community, the Cloud Native Computing Foundation, and the Linux Foundation to create the foundation layers for agentic infrastructure in open source:
- kagent — a CNCF project that extends Kubernetes into a context-aware runtime..
- kmcp — a full lifecycle toolkit for creating and deploying MCP servers.
- agentgateway — the industry’s first agent-native data plane, supporting comprehensive AI connectivity across AI gateway, inference gateway, and agentic protocols (MCP, A2A).
Individually, these projects address specific gaps in using Kubernetes for agentic infrastructure. Together, they form the backbone of the industry’s first enterprise-ready agentic infrastructure platform in kagent enterprise. Kagent enterprise builds on this foundation with the security, governance, and enterprise-scale controls required to bring agentic infrastructure into production.
The future of agentic infrastructure is here - and it’s enterprise-ready. Get started today:
- Follow our getting started guide to experiment with AI agents in your Kubernetes cluster.
- Get involved by contributing via GitHub.
- Join the discussion in discord.
%20a%20Bad%20Idea.png)
%20For%20More%20Dependable%20Humans.png)
