OpenClaw — Optional Agent Backend

OpenClaw is an optional agent implementation behind the OpenKubes Agent Interface Contract. The repository describes a token-authenticated, stateless deployment, integrated with Open WebUI as an agent-backed model.

Architecture and boundaries

• Open WebUI provides the chat-facing interface; agent runtimes remain replaceable behind a defined contract.
• Platform diagnostics are accessed through the bounded, read-only diagnostics/MCP interface, not through inherited administrator credentials. kagent & Platform Diagnostics explains the first diagnostics provider profile.
• Agent integrations must respect explicit authorization, tool boundaries and evidence collection.
• Experimental PoC work and roadmap entries must not be interpreted as a generally available autonomous production operations service.

Source-owned documentation

• OpenClaw component README — status, architecture, security controls and component-specific deployment guidance.
• Agentic AI implementation guideline.
• ADR-Platform-015 — Agentic AI.
• ADR-Platform-021 — Read-Only Platform Diagnostics.
• Agentic Operations — broader reference architecture and governance.

Operational note: Only run agent deployments after reading the current component README and validating target-cluster access and security requirements.