k8sgpt

MCP.Pizza Chef: k8sgpt-ai

Years of hard-won operations experience are baked into its checks, so instead of a wall of logs you get a short list of what is actually wrong and what to try next. Switch on its server mode and Claude Desktop or another assistant can run those checks for you, look at overall health, and dig into a single namespace. Install it with Homebrew or a package, then point it at the AI service you prefer — OpenAI, Azure, Amazon Bedrock, Google Gemini, Cohere or a local model, with your own key.

Coding

Use This MCP server To

Find out why a service keeps restarting Get a plain-English summary of cluster health Check one namespace before a release Ask what to fix first when several things break Review suggested fixes before I run anything

README

Text changing depending on mode. Light: 'So light!' Dark: 'So dark!'

GitHub code size in bytes GitHub Workflow Status GitHub release (latest by date) GitHub last commit (branch)

k8sgpt is a tool for scanning your Kubernetes clusters, diagnosing, and triaging issues in simple English.

It has SRE experience codified into its analyzers and helps to pull out the most relevant information to enrich it with AI.

Out of the box integration with OpenAI, Azure, Cohere, Amazon Bedrock, Google Gemini and local models.

k8sgpt FAQ

Which apps does it work in?

Claude Desktop and other MCP clients — start it in server mode, then add it to the app's settings.

Do I need a key?

Yes — you pick an AI service such as OpenAI, Azure, Amazon Bedrock, Google Gemini or Cohere, or run a local model instead.

Can I use this to work out why one namespace keeps failing?

Yes — ask about that namespace and you get the problems it found plus suggested fixes.

Will it change my cluster?

Its focus is scanning, explaining and suggesting, so read anything it proposes before acting on it.

How hard is setup?

Install with Homebrew or a package, sign in to your AI service, then start the server and paste a small settings block.

Can it watch a cluster continuously?

There is a separate operator you can install inside the cluster for ongoing checks and alerts.