Conversations
2 min read
Solo.io: AI agents as microservices, but more
Lin Sun of Solo.io on the infrastructure that runs AI agents in production: agentgateway, kagent and Agent Substrate on Kubernetes, and how to keep up.
Lin Sun works in open source at Solo.io. Luca Berton talked to her at Open Source Summit Europe in Prague on 8 October 2026. She speaks all over the world: Amsterdam in March, Amsterdam again in September.
Ten years of cloud native
Lin started contributing to CNCF projects about 10 years ago, when Kubernetes had just started; her first contribution to Kubernetes was in January 2016. She has been a KubeCon speaker since 2017 in Berlin, and has attended and spoken at every KubeCon since, in Europe and North America. At KubeCon in Amsterdam she was one of the conference co-chairs, “So I was able to do a fun keynote for everybody.”
Infrastructure for AI agents
Her focus is the infrastructure that lets AI agents deploy and scale, and keeps them secure, with the right governance and observability, in production. One project she contributes to is agentgateway: it secures, observes and governs the communication of AI agents with MCP servers, with large language models and with other agents.
Microservices, but more
“So you can think about AI agents as microservices, but more, right?” The microservice world had to solve connectivity: how to connect, secure and observe microservices. AI agents have the same problem, with a lot more focus on security, identity and policies, “because we don’t trust the AI agents”, and because agents can work on behalf of you, me or other team members. That is also why a sandbox environment for AI agents is becoming super important.
kagent and Agent Substrate
kagent makes it easy to deploy AI agents in Kubernetes. A hot new project, Agent Substrate, adds a layer of scheduling on top of Kubernetes, to place AI agents in a Kubernetes cluster, on top of pods, efficiently and at large scale. Lin has started contributing to it as well.
Keeping up
How does she keep up with new protocols and techniques? “It is hard. It is super hard.” She spends probably around 50% of her time learning, because the space is evolving so fast. Another good chunk goes into building on these technologies, and another into fixing, in the projects themselves, the problems she meets while using them, “so that way I can not only be a consumer of these open source projects, but also be a contributor”.
Join the community meetings
agentgateway has a community meeting every Wednesday; the information is on agentgateway.dev. kagent.dev has its community information too, with a meeting she believes is every Tuesday. Agent Substrate meets every Thursday, and it is probably the largest community she has ever taken part in: “Every time I join the community meeting, it’s like 50 people plus.”
More
The episode with Lin Sun is edited and goes out on YouTube first, then on Apple Podcasts, Spotify and Amazon Music.
Built something the rest of us run on? Being a guest is free. Take the other chair
The episodes in this article
All episodes →-
Episode 84 · Coming
Lin Sun: AI agents as microservices, but more