home / mcp / inspektor gadget mcp server

Inspektor Gadget MCP Server

Debug your Container and Kubernetes workloads with an AI interface

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "inspektor-gadget-ig-mcp-server": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "--volume",
        "ig-mcp-cache:/root/.cache/ig-mcp-server",
        "--mount",
        "type=bind,src=${env:HOME}/.kube/config,dst=/kubeconfig",
        "ghcr.io/inspektor-gadget/ig-mcp-server:latest",
        "-gadget-discoverer=artifacthub"
      ],
      "env": {
        "IG_MCP_CACHE": "$HOME/.cache/ig-mcp-server"
      }
    }
  }
}

You run the Inspektor Gadget MCP Server to enable AI-powered debugging and inspection for Kubernetes clusters from your MCP client. It provides automated gadget discovery, one-click deployment, and intelligent analysis to help you troubleshoot and monitor clusters directly from your development environment.

How to use

You connect to the MCP server from your MCP-enabled client. The server exposes gadget-based tools as MCP endpoints. Use the available gadgets to perform troubleshooting tasks such as DNS tracing or process/socket snapshots, and review summarized results. Gadget discovery can be automatic via Artifact Hub or manual by selecting specific gadgets to load.

How to install

Prerequisites you need before installation:

- Docker installed on your workstation or CI environment.

- A valid kubeconfig file that can access your Kubernetes cluster.

Install using Artifact Hub discovery (Docker method)

code --add-mcp '{
  "name": "inspektor-gadget",
  "command": "docker",
  "args": [
    "run",
    "-i",
    "--rm",
    "--volume",
    "ig-mcp-cache:/root/.cache/ig-mcp-server",
    "--mount",
    "type=bind,src=${env:HOME}/.kube/config,dst=/kubeconfig",
    "ghcr.io/inspektor-gadget/ig-mcp-server:latest",
    "-gadget-discoverer=artifacthub"
  ]
}'

Install using explicit gadgets (Docker method)

code --add-mcp '{
  "name": "inspektor-gadget",
  "command": "docker",
  "args": [
    "run",
    "-i",
    "--rm",
    "--volume",
    "ig-mcp-cache:/root/.cache/ig-mcp-server",
    "--mount",
    "type=bind,src=${env:HOME}/.kube/config,dst=/kubeconfig",
    "ghcr.io/inspektor-gadget/ig-mcp-server:latest",
    "-gadget-images=trace_dns:latest,trace_tcp:latest,snapshot_process:latest,snapshot_socket:latest"
  ]
}'

Binary installation options

You can download and install a prebuilt binary for your platform, then add it to your MCP configuration.

# Linux
MCP_VERSION=$(curl -s https://api.github.com/repos/inspektor-gadget/ig-mcp-server/releases/latest | jq -r .tag_name)
MCP_ARCH=amd64
curl -sL https://github.com/inspektor-gadget/ig-mcp-server/releases/download/${MCP_VERSION}/ig-mcp-server-linux-${MCP_ARCH}.tar.gz | sudo tar -C /usr/local/bin -xzf - ig-mcp-server

# macOS
MCP_VERSION=$(curl -s https://api.github.com/repos/inspektor-gadget/ig-mcp-server/releases/latest | jq -r .tag_name)
MCP_ARCH=arm64
curl -sL https://github.com/inspektor-gadget/ig-mcp-server/releases/download/${MCP_VERSION}/ig-mcp-server-darwin-${MCP_ARCH}.tar.gz | sudo tar -C /usr/local/bin -xzf - ig-mcp-server

# Windows (PowerShell)
$MCP_VERSION = (curl.exe -s https://api.github.com/repos/inspektor-gadget/ig-mcp-server/releases/latest | ConvertFrom-Json).tag_name
$MCP_ARCH = "amd64"
curl.exe -L "https://github.com/inspektor-gadget/ig-mcp-server/releases/download/$MCP_VERSION/ig-mcp-server-windows-$MCP_ARCH.tar.gz" -o "ig-mcp-server.tar.gz"
$destPath = "C:\\Program Files\\ig-mcp-server"
if (-Not (Test-Path $destPath -PathType Container)) { mkdir $destPath}
tar.exe -xzf "ig-mcp-server.tar.gz" -C "$destPath"
rm ig-mcp-server.tar.gz
Write-Host "✅ Extracted to $destPath"
Write-Host "👉 Please add '$destPath' to your PATH environment variable manually."

Finalize MCP configuration in your client

After configuring the MCP server, you can start using AI commands and gadget-based tools in your client. The server supports automatic gadget discovery and explicit gadget loading, allowing you to tailor the available tools to your needs.

Configuration options and notes

- Gadget discovery can be automatic via Artifact Hub or manual by specifying gadgets directly.

- You can limit the tools available by choosing the gadget-discoverer or gadget-images options.

- The server requires read-only access to your kubeconfig and needs network access for Artifact Hub discovery.

Security considerations

- Ensure your kubeconfig permissions are restricted to what is necessary.

- Review network access policies for Artifact Hub discovery and gadget downloads.

Resources and support

Explore related tooling and community channels for support, troubleshooting, and updates.

Available tools

ig_deploy

Manage deployment of Inspektor Gadget on the target Kubernetes cluster

ig_gadgets

Manage the lifecycle of running gadgets (stop/list/get-results)

gadget_trace_dns

Trace DNS traffic gadget for troubleshooting DNS behavior in the cluster

gadget_trace_tcp

Trace TCP connections gadget for monitoring network activity in the cluster

gadget_snapshot_process

Snapshot process state for debugging pod behavior

gadget_snapshot_socket

Snapshot sockets for in-depth process/socket analysis