Agent Substrate, with Tim Hockin and Brandon Royal
Source: YouTube · The Kubernetes Podcast from Google · published Sep 3, 2026 · 47:40
BLUF: Agent Substrate is a new open-source project designed to run AI agents at scale on Kubernetes by introducing primitives like suspend/resume and multiplexing to address the high idleness and isolation needs of agent workloads 2:15.
Key Takeaways:
• Solving Agent Idleness: Agent workloads are often idle over 90% of the time while waiting for LLM responses or human input, leading to wasted compute resources. Agent Substrate addresses this with suspend and resume capabilities, allowing systems to snapshot and store agent states rather than keeping pods idle 3:30.
• Beyond Standard Kubernetes Primitives: Traditional Kubernetes pods are too slow to spin up (seconds) for agents requiring sub-second responsiveness. Substrate decouples the "actor" (agent session) from the "worker" (pod), enabling rapid context switching and better density through multiplexing 6:20.
• Relationship with Agent Sandbox: While Agent Sandbox optimizes the lifecycle of Kubernetes pods for agents, Agent Substrate introduces a new layer of abstraction. It allows agents to move between different underlying sandbox technologies (like gVisor or MicroVMs) without being tied to a specific pod's lifecycle 8:40.
• Network and Identity Challenges: The project utilizes a router (AtNet) to manage traffic, automatically waking suspended actors and handling identity delegation. This ensures agents have precise, dynamic access controls without being bound to static pod IPs 21:30.
• Future Roadmap: The project is actively developing core use cases for productivity and coding agents, aiming for production readiness by KubeCon. It seeks to bridge the gap between Kubernetes' maturity and the rapid pace of AI agent development 33:10.
Agent Substrate represents a significant shift in how we infrastructure AI agents, moving from static pod management to dynamic, state-aware execution models that leverage Kubernetes' strengths while overcoming its limitations for agentic workloads.
Sources:
- 2:15 Introduction to Agent Substrate and its purpose.
- 3:30 Discussion on agent idleness and resource waste.
- 6:20 Limitations of Kubernetes startup times for agents.
- 8:40 Distinction between Agent Sandbox and Agent Substrate.
- 21:30 Explanation of the AtNet router and automatic wakeups.
- 33:10 Project roadmap and production goals.
Generate CPE Credits
Generate a professional CPE document from this video's transcript.
Estimated credit: 1 CPE hours
Estimate uses the video runtime (1 hour ≈ 1 CPE, rounded to the nearest 0.5, minimum 0.5, maximum 2.0). The final amount can be lower after review, never higher.
Topic: Cloud Security. Commonly maps to: Security Architecture and Engineering, Communication and Network Security. Exact CISSP domains are assigned during generation.
Free account. One generation at a time, with a daily limit.
CPEBuddy is independent and not affiliated with or endorsed by ISC2, ISACA, or any certification body. Exports are formatted for common CPE submissions; acceptance is at your certification body's discretion.
Transcript Preview
First 800 characters of the transcript
Hello and welcome to the Kubernetes podcast from Google. I'm your host Kaslyn Fields. [music] In this episode, we sat down with two legends in the cloudnative community, Tim Hawin, one of the original founders of Kubernetes, and Brandon Royal, who drove the upstream work on many projects, including the agent sandbox. We spoke about agent substrate, a new open-source project designed to run agents at scale on Kubernetes. But first, let's get to the news. Kubernetes 1.37 is out now. If you haven't yet, check out the release blog and listen in to our latest episode where we interviewed the release lead, Depetra Wat. Kubernetes 1.37's release signals the nearing end of maintenance for version 1.34. Support for 1.34 is scheduled to end on October 27th, 2026, meaning the release will no longer r…