This keynote explores how Kubernetes should evolve from a cloud-native platform into an agent-native runtime, able to supporting rising AI and agentic workloads at scale. Offered by Solo.io Founder & CEO Idit Levine and Chief Product Officer Keith Babo, the session highlights why many agentic AI pilots stall earlier than reaching manufacturing, regardless of sturdy developer capabilities.
The core challenge, they clarify, will not be constructing brokers – it’s working them securely, observably, and reliably in manufacturing environments.
The audio system element how at present’s networking infrastructure, together with Envoy-based proxies, falls brief when coping with agentic site visitors that requires deep inspection of request our bodies for context extraction, MCP communication, and LLM token accounting. To deal with this, Solo.io constructed a brand new Rust-based knowledge airplane optimized for AI workloads and open-sourced it as agent-gateway, delivering superior context-aware routing and excessive throughput.
Past connectivity, they underline the necessity for Kubernetes itself to develop into a context-aware runtime for managing brokers, instruments, and stateful workflows. The open-sourced kagent mission extends Kubernetes to deal with brokers as first-class workloads with full id, coverage and observability. To finish the manufacturing stack, the audio system introduce agent-registry, a lifecycle administration platform for locating, curating and securing AI artifacts corresponding to instruments, brokers and fashions.
Collectively, agent-gateway, kagent, and agent-registry type the inspiration of a brand new operational mannequin – context engineering – bringing platform engineering rules to the world of agentic AI. The keynote concludes with an outline of how rising requirements like MCP and Anthropic Agent Expertise combine into this ecosystem, enabling composable and shareable workflows that may run securely throughout Kubernetes.
This session delivers deep insights for platform engineers, AI infrastructure architects, and Kubernetes practitioners making ready to help next-generation agentic methods.
Right here is an Government Insights FAQ about this keynote’s core themes:
How does context engineering lengthen Kubernetes for agentic workloads?
Context engineering introduces new runtime capabilities – id, observability, and context-aware networking – so Kubernetes can help brokers that require reminiscence, coverage, and human-in-the-loop workflows.
Why can’t current Layer 7 proxies deal with AI and MCP site visitors successfully?
Conventional proxies solely manipulate headers, however agentic workflows embed vital directions and context throughout the physique. MCP’s JSON-RPC over HTTP makes deep physique inspection important, requiring new proxy architectures.
What position does kagent play in operationalizing brokers?
kagent elevates brokers and instruments to first-class Kubernetes workloads, enabling safe identities, coverage enforcement, and end-to-end observability throughout various agent frameworks and runtimes.
Why is agent-registry turning into a vital element of AI platforms?
Organizations want a managed technique to publish, uncover, model, and govern AI artifacts. Agent-registry gives lifecycle administration throughout non-public and public registries, supporting safety and compliance.
How do Anthropic Agent Expertise change the agent growth mannequin?
Expertise introduce reusable, composable, deterministic workflows. Mixed with MCP instruments and native instruments, they permit brokers to jot down, execute, and share capabilities—requiring sandboxing and lifecycle controls constructed into the platform.
