AI-powered Kubernetes cluster analyzer and remediation tool. Built-in analyzers scan pods, services, deployments, ingresses, and events for misconfigurations and failures, providing plain-English explanations via multiple AI backends (OpenAI, Azure, Bedrock, local models). Operator mode enables continuous in-cluster monitoring. Experimental auto-remediation patches supported resources. MCP server exposes cluster operations as tools for AI assistants.
K8sGPT is an open-source CLI and Kubernetes Operator that runs targeted analyzers across cluster resources and uses a connected LLM to explain failures in plain English. Built-in analyzers cover pods, services, deployments, replicasets, statefulsets, ingresses, and Kubernetes events. Each analyzer fetches relevant resource state and passes a structured prompt to the configured AI backend.
Supported backends: OpenAI, Azure OpenAI, AWS Bedrock, Google Gemini, Ollama (local models), and others via provider plugin. Operator mode runs analyzers continuously inside the cluster, surfacing findings as Kubernetes custom resources for integration with monitoring pipelines. The MCP server exposes analyzers as tools for AI coding agents.
Experimental auto-remediation patches a narrow set of resource types. CNCF Sandbox project under Apache 2.0.
Key Features
Targeted Kubernetes analyzers: structured inspection of pods, services, deployments, statefulsets, replicasets, ingresses, and cluster events with LLM-generated plain-English failure explanations
Multiple AI backend support: OpenAI, Azure OpenAI, AWS Bedrock, Google Gemini, Anthropic, and Ollama — users supply their own API key or run a local model for sensitive clusters
Kubernetes Operator mode: runs analyzers continuously inside the cluster and surfaces findings as Kubernetes custom resources for integration with existing monitoring and alerting pipelines
MCP server: exposes K8sGPT analyzers as tools for Claude Code and other MCP-compatible AI agents, enabling cluster investigation as part of agentic automation workflows
Experimental auto-remediation: applies fixes for a defined set of resource misconfiguration patterns — not recommended for production use without testing and guardrails
CNCF Sandbox project under Apache 2.0: no vendor dependency, self-hostable, and extensible via provider and analyzer plugin interfaces
Integrations
3 total
orchestration
Kubernetes
messaging
Slack
monitoring
Prometheus
Pricing
1 tier
Open Source
Free
Free tier: Fully free and open source (CNCF Sandbox, Apache 2.0). No commercial tier.
