OK AI Roadmap
From trusted answers to governed operations. OpenKubes is developing OK AI in three deliberately separated phases: documentation assistance, read-only diagnostics, and optional, policy-governed operations.
Phase 01
Documentation Assistant
Ask questions about OpenKubes architecture, installation, and configuration using allowlisted public documentation, READMEs, and tutorials.
Read-only answers. No cluster access or command execution.
Explore the chat UI prototype โPhase 02
Read-only Diagnostics
Analyze user-supplied, redacted logs and Kubernetes manifests, with future support for explicitly scoped, read-only diagnostic data.
Investigate and explain. Never change cluster state.
Read about platform diagnostics โPhase 03
Agentic Operations
Authenticated, policy-governed platform operations with explicit authorization, a bounded runner, and independently verifiable evidence.
Agents propose. OpenKubes authorizes and executes. Evidence proves.
Explore the reference architecture โRoadmap, not a release announcement
Phase 1 has passed a private lab integration check for one documentation question, not a public beta or a semantic-grounding certification. Installation-command handling and public security, quality, and release checks are still in progress. Phases 2 and 3 describe optional directions, not generally available features. OpenKubes remains functional without AI agents.
One architecture, clear trust boundaries
AI thinks. OpenKubes executes. Evidence proves. Each phase adds capabilities without giving an AI agent unrestricted access to Kubernetes. In Phase 3, platform state changes must pass through independent policy, approval, and execution contracts; the agent never receives direct Kubernetes mutation authority.
For the publicly documented execution boundaries, see the Agentic Operations Reference Architecture.