AI
Why KubeStellar Console Bans Human Pull Requests
KubeStellar Console is built on a strict rule: no human PRs land in the repo. All code changes go through GitHub agentic workflows, a feature Anderson noted most attendees had not yet heard of. Those workflows run Claude, Gemini, and Codex underneath, along with GitHub Copilot, which itself pulls from the same models. The constraint is deliberate. Ten years of maintaining Kubernetes clusters informed the decision to remove manual kubectl work from the loop entirely.
Multi-Cluster Access From a Single kubeconfig
The console reads whatever clusters are already defined in your kubeconfig and deduplicates entries so the same cluster does not appear twice. Permissions stay exactly as configured. If your API grants read-only access, the console surfaces read-only access and nothing more. No additional credentials are required. This design means teams with dozens of clusters can see all of them in one place without renegotiating access controls or maintaining a separate inventory.
AI Missions Replace Individual kubectl Commands
Instead of writing prompts from scratch each session and hoping you saved the right markdown file, the console stores and shares missions. A mission can be as specific as deploying Kyverno to all production clusters, or as broad as running an ISO 27001 compliance audit across the fleet. Anderson asked the room how long that audit normally takes. The answer from the audience was “a long time.” The console returns the result in seconds, and the mission stays always-on for continuous monitoring.
185 CNCF Projects, One Dashboard
Support spans 185 of the CNCF’s current 210 projects. The main dashboard shows roughly 30 sidebar panels, stat blocks that update dynamically by pod or deployment count, and cards for drilling into specifics. Anderson is deliberately holding the panel count near 30 to avoid overwhelming new users. From a single interface, teams can compose missions that combine Argo, OPA, Kyverno, and Trestle rather than running each tool separately. The console installs in 60 seconds or less via a curl command available at console.kubstellar.io in demo mode.
Community Leaderboard and White-Label Option
Anderson closed with a challenge to the roughly 200 people in the room: open one legitimate bug. Any bug filed will be implemented by the end of the week, handled by the AI pipeline rather than a human review queue. A public leaderboard tracks contributions at kubebe.io/leaderboard. For enterprises wanting to steer the project toward internal use cases, the console supports white labeling. Anderson invited anyone interested to speak with the team directly at the conference.
Notable Quotes
this is the world’s first fully integrated and automated Kubernetes management and orchestration solution Andy Anderson · ▶ 00:28
We don’t allow human PRs in this repo. So you can all gasp now. Andy Anderson · ▶ 00:35
how many of you have been asked to run an ISO27001 audit compliance recently how about how long did that take you Andy Anderson · ▶ 00:45
I challenge you to open a legitimate bug. It will be implemented by the end of the week. Andy Anderson · ▶ 05:50
Key Takeaways
- KubeStellar Console blocks human pull requests; Claude, Gemini, and Codex handle all code changes.
- One kubeconfig connects all clusters with no permission escalation beyond what already exists.
- AI missions cover 185 CNCF projects and can run ISO 27001 audits in seconds rather than days.
About the Speaker(s)
Andy Anderson is a Software Architect at IBM with nearly 10 years working on Kubernetes clusters. He leads research and engineering teams focused on hybrid cloud and Kubernetes-based platform development, combining GitOps practices, containerization, and cloud computing to support platform engineering at scale.