# Red Hat OpenShift Lightspeed 2 adds autonomous Kubernetes troubleshooting with human approval gates

> **Key takeaway:** OpenShift Lightspeed 2 introduces two operational modes for cluster troubleshooting: an interactive chat with live cluster queries and read-write tools, plus a Technology Preview autonomous agent that investigates problems in an isolated pod with explicit human approval for any changes.

**Published:** 2026-11-13T02:02:27.559393+00:00 
**Source:** Red Hat Blog 
**Category:** distro-news 
**Canonical URL:** https://fosswire.org/news/red-hat-openshift-lightspeed-2-adds-autonomous-kubernetes-tr.html 

## Summary
Red Hat has released OpenShift Lightspeed 2, a major evolution of its console-embedded AI assistant for OpenShift clusters. The tool advances from answering documentation questions to actively diagnosing and resolving Kubernetes problems through two new modes. The generally available troubleshooting chat performs live cluster investigation with read-write capabilities, pausing for explicit user approval before making changes. A Technology Preview autonomous mode spins up an isolated execution environment to independently investigate issues and propose fixes. Both modes maintain human-in-the-loop control, with the autonomous agent running in a protected pod with its own workspace and OpenShift CLI access.

The release builds on capabilities introduced across the 1.x series, including a Model Context Protocol server for cluster read access (1.1), support for Amazon Bedrock and Google Vertex AI providers, automatic ROSA cluster detection, OpenTelemetry tracing, and configurable TLS profiles. New integrations include Perses Dashboards for dynamic metric visualization within conversations. Future directions under consideration include institutional memory across runs, CLI and API access beyond the console, and integration with OpenShift GitOps infrastructure-as-code workflows.

## Operational modes and cluster integration in Lightspeed 2
- **Troubleshooting chat (GA):** Performs live cluster queries including pod listing, event reading, resource limit checks, and node condition inspection; supports mid-investigation redirection and pauses for explicit approval before any mutating operation.
- **Autonomous agentic runs (Technology Preview):** Creates an isolated pod with dedicated workspace and OpenShift CLI access to independently build investigation plans, execute checks, and adjust approach until root cause identification; returns proposed fixes for human approval.
- **Read-write tool expansion:** Extends the 1.x MCP server's read-only cluster access to include write capabilities, enabling the assistant to implement fixes directly rather than only verifying and instructing.
- **Perses Dashboards integration:** Generates dynamic metric visualizations rendered directly within the chat conversation for cluster metrics requested by users.
- **Provider and platform support:** Maintains compatibility with multiple model backends including Amazon Bedrock, Google Vertex AI, and OpenAI; adds automatic detection for Red Hat OpenShift Service on AWS (ROSA) clusters to reduce configuration overhead.

## Why it matters
For Kubernetes platform engineers and SREs, Lightspeed 2 represents a concrete step toward operational AI agents that respect enterprise safety requirements through mandatory approval gates rather than autonomous action. The isolated pod architecture for agentic runs addresses multi-tenant security concerns that have slowed AI adoption in production infrastructure tooling. The planned CLI and API expansions signal Red Hat's intent to make this operational model available to broader automation workflows, potentially influencing how other cloud-native vendors design human-in-the-loop AI for infrastructure management.