Nice work. Agents are quickly becoming infra, not just local dev tools.
Keeping credentials inside the tool container feels much closer to how this should work.
But in a lot of teams the starting point is not “we need an agent platform”. It is the CEO saying: "we give everyone Claude Code accounts, go use them ASAP!!!".
Do you expect teams to jump from local Claude/Codex/Cursor usage straight into a k8s-native runtime?
Congrats on the launch! I’m happy to see OpenZiti in your networking stack alongside Istio. The combo of sandbox-per-agent and MCPs-in-separate-containers is exactly the kind of deployment that benefits from identity-based, zero-listening-port connectivity between components.
If you're up for it, we'd love to have someone from Agyn on Ziti TV (our livestream) to talk about how you're using OpenZiti and what you're building on the agent-infra side. Drop me a line if you're interested — joey.devilla@netfoundry.io.
Disclosure: As my email address implies, I work at NetFoundry, which created and sponsors OpenZiti. I’m their Developer Advocate and it’s my job to promote OpenZiti (and especially for AI tools and platforms), but either way, I really like what you're shipping here.
But in a lot of teams the starting point is not “we need an agent platform”. It is the CEO saying: "we give everyone Claude Code accounts, go use them ASAP!!!".
Do you expect teams to jump from local Claude/Codex/Cursor usage straight into a k8s-native runtime?