[ANNOUNCE] Agent Sandbox Subproject & Release v0.1.1

[Janet Kuo]

Hi All,

I’m excited to formally announce the Agent Sandbox subproject! As some of you may have heard during the announcement at KubeCon NA 2025, this project enables easy management of isolated, stateful, singleton workloads on Kubernetes, ideal for use cases like AI agent runtimes.

🚀 Release Highlights

We previously shipped v0.1.0, our first official release, introduced the core declarative APIs (Sandbox) and extensions (SandboxTemplate, SandboxClaim, SandboxWarmPool) for faster provisioning.

We also just released v0.1.1, which brings significant improvements to stability and observability:

• New Documentation Site: We launched a dedicated site at agent-sandbox.sigs.k8s.io.
• OpenTelemetry Support: Added optional tracing to both the Python client and Controllers.
• Enhanced Features: Includes configurable shutdown policies and better extension management.
• Critical Fixes: Major refactors for full gVisor (runsc) compatibility in the Python SDK and improved WarmPool reliability.
• New Examples: Check out our new examples, including "Gemini Computer Use" and "Moltbot".

Getting Started

You can install the core components and extensions directly from the release page:

kubectl apply -f https://github.com/kubernetes-sigs/agent-sandbox/releases/download/v0.1.1/manifest.yaml
kubectl apply -f https://github.com/kubernetes-sigs/agent-sandbox/releases/download/v0.1.1/extensions.yaml

For more details, check out our GitHub Repository.

Get Involved

We host biweekly community meetings on Mondays 9:30-10am PST / 12:30-1:00pm EST / 6:30-7:00pm CET (alternating with the SIG Apps meeting). We’d love to see you there!

Cheers, 

Janet Kuo, on behalf of SIG Apps Leads