Getting Started

Connect any MCP-compatible AI assistant to Kubeshark and query your Kubernetes network traffic using natural language.

Quick Setup

Claude Code (Terminal)

claude mcp add kubeshark -- kubeshark mcp

Claude Desktop

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "kubeshark": {
      "command": "kubeshark",
      "args": ["mcp"]
    }
  }
}

Cursor / VS Code

Configure in your editor’s MCP settings with the same command and arguments.

Connection Modes

ModeUse When
URL ModeKubeshark is already running and accessible
Proxy ModeLet the CLI proxy into your cluster via kubectl
# Default - uses current kubeconfig
kubeshark mcp

# URL mode - connect to a running instance
kubeshark mcp --url https://kubeshark.example.com

# Explicit kubeconfig
kubeshark mcp --kubeconfig ~/.kube/config

Your First Query

Once connected, try:

“What services are running in my cluster?”

“Show me any HTTP 500 errors in the last hour.”

“Which services communicate with the payment service?”

The AI will use Kubeshark’s MCP tools to query your traffic and return insights.

Conversational Debugging

Ask debugging questions naturally instead of writing queries:

“Is my user-service receiving requests from the API gateway? Show me the last 10 requests.”

“What exactly is the notification-service sending to the email-provider? Show me the request body.”

“The frontend says it’s sending the right headers, but the backend disagrees. Show me what’s actually in the HTTP request.”

“Find all 4xx errors returned by my service in the last 30 minutes. What requests caused them?”

The AI can show actual request/response payloads, verify headers, find errors and their corresponding requests, compare what’s sent vs. what’s received, and show connection-level issues like TCP resets and timeouts.

Autonomous Development

AI coding assistants can write and deploy code, but they lack visibility into how that code actually behaves in Kubernetes. Kubeshark closes this gap by providing real-time network feedback, enabling AI tools to identify issues and fix them — deploy, verify, and fix in one autonomous loop.

“Deploy my changes and verify the new /api/orders endpoint works correctly.”

“Run the integration tests and use Kubeshark to verify the API calls are correct — check payloads, headers, and downstream calls.”

“I changed the retry logic. Deploy and verify that failed requests are retried exactly 3 times with exponential backoff.”

Kubeshark provides feedback that logs, metrics, and test assertions miss — malformed payloads, unexpected retries, missing headers, N+1 queries, serialization bugs, and connection pooling issues.

What’s Next