How addithya-sri/ai-assisted-kubernetes-incident-diagnosis-using-model-context-protocol-mcp- works
This repository is a runbook and configuration prototype for AI-assisted, read-only Kubernetes incident diagnosis—not an implementation of the MCP server itself. The documented workflow is to create a disposable kind cluster with a healthy nginx workload, connect an AI assistant through MCP, deliberately break the workload, ask the assistant to investigate and explain the cause, then revert and capture the demo. The MCP server supplies the Kubernetes inspection tools and is the documented read-only gate. Runtime wiring is based on the README; source excerpts were not available.
Main entry points: 00-setup.sh Cluster setup · 03-break-it-scenarios.sh Break and revert · 99-teardown.sh Cluster teardown · 01-nginx-test-deployment.yaml nginx-demo workload
Components 12
Assistant interaction
- AI assistant
- Diagnosis prompts
prompts/diagnosis-prompts.md - Demo transcript
docs/transcript-template.md
MCP diagnosis
- Claude MCP config
.mcp.json - Cursor MCP config
.cursor/mcp.json - Kubernetes MCP server
Cluster demo
- kind cluster
- nginx-demo workload
manifests/01-nginx-test-deployment.yaml - Cluster setup
scripts/00-setup.sh - Break and revert
scripts/03-break-it-scenarios.sh - Cluster teardown
scripts/99-teardown.sh
Other components
- Engineer
Connections 17
- Engineer → Cluster setup: starts cluster
- Cluster setup → kind cluster: creates baseline
- Cluster setup → nginx-demo workload: deploys workload
- Engineer → Break and revert: selects scenario
- Break and revert → nginx-demo workload: breaks or reverts
- Engineer → AI assistant: asks diagnosis
- Diagnosis prompts → Engineer: guides questions
- AI assistant → Claude MCP config: reads connection
- AI assistant → Cursor MCP config: reads connection
- AI assistant → Kubernetes MCP server: calls tools
- Kubernetes MCP server → kind cluster: reads state
- kind cluster → Kubernetes MCP server: returns status, events, logs
- Kubernetes MCP server → AI assistant: returns findings
- AI assistant → Engineer: explains root cause
- Engineer → Demo transcript: captures demo
- Engineer → Cluster teardown: cleans up
- Cluster teardown → kind cluster: removes cluster