Skip to main content
Glama

mcp-devops-server

An MCP (Model Context Protocol) server that gives AI agents access to DevOps tools — kubectl, terraform, helm, docker, and AWS cost analysis.

Works with Claude Desktop, Cursor, GitHub Copilot, and any MCP-compatible client.

Tools

Tool

Description

kubectl_get

Get Kubernetes resources

kubectl_describe

Describe a Kubernetes resource

kubectl_logs

Get pod logs

terraform_plan

Run terraform plan

terraform_show

Show terraform state

helm_list

List Helm releases

helm_diff

Preview Helm upgrade changes

docker_ps

List Docker containers

docker_logs

Get container logs

aws_cost_summary

Get AWS cost summary (last 7/30 days)

aws_resource_inventory

List AWS resources by service

Related MCP server: K8s MCP Server

Installation

Using npx (no install)

npx @basel5001/mcp-devops-server

Global install

npm install -g @basel5001/mcp-devops-server
mcp-devops-server

From source

git clone https://github.com/basel5001/mcp-devops-server.git
cd mcp-devops-server
npm install
npm run build
npm start

Configuration

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "devops": {
      "command": "npx",
      "args": ["-y", "@basel5001/mcp-devops-server"]
    }
  }
}

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "devops": {
      "command": "npx",
      "args": ["-y", "@basel5001/mcp-devops-server"]
    }
  }
}

Security Considerations

  • Command allowlist: Only kubectl, terraform, helm, docker, and aws commands can be executed.

  • Argument sanitization: Shell metacharacters (;, &, |, `, $) and path traversal (..) are rejected.

  • No shell execution: Commands run via execFile (not exec), preventing shell injection.

  • Timeouts: All commands have configurable timeouts (default 30s) to prevent hanging.

  • Read-only by design: No destructive operations (delete, apply, destroy) are exposed.

Development

npm install
npm run dev    # watch mode
npm run build  # compile
npm run lint   # lint

License

MIT

Available Tools

11 tools
aws_cost_summaryC

Get AWS cost summary

Inputs: days, granularity.

aws_resource_inventoryC

List AWS resources by service

Inputs: service (required), region.

docker_logsB

Get Docker container logs

Inputs: container (required), tail, since.

docker_psC

List Docker containers

Inputs: all, filter.

helm_diffB

Preview Helm upgrade changes

Inputs: chart (required), release (required), values, namespace.

helm_listC

List Helm releases

Inputs: namespace, allNamespaces.

kubectl_describeC

Describe a Kubernetes resource

Inputs: name (required), resource (required), namespace.

kubectl_getC

Get Kubernetes resources

Inputs: resource (required), name, output, namespace.

kubectl_logsC

Get logs from a Kubernetes pod

Inputs: pod (required), tail, since, container, +1 more.

terraform_planC

Run terraform plan

Inputs: directory (required), vars.

terraform_showC

Show terraform state

Inputs: directory (required), json.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3/5.0

    Scored across 11 tools

    Disambiguation5/5

    Every tool uses a platform prefix that clearly separates Kubernetes, Terraform, Helm, Docker, and AWS concerns. Within each platform, the actions target distinct resources or operations, so no two tools are easily confused.

    Naming Consistency4/5

    Tool names mostly follow a predictable <platform>_<command-or-action> pattern, such as kubectl_get, terraform_show, and helm_list. The two AWS tools use noun phrases instead of verbs, which is a minor deviation from the otherwise consistent style.

    Tool Count5/5

    With 11 tools, the server covers a broad DevOps range without feeling bloated. Each tool represents a distinct and useful operation for infrastructure inspection and planning.

    Completeness2/5

    The toolset is heavily oriented toward read-only and planning operations: logs, describe, list, plan, show, and diff. There are no apply, deploy, delete, or mutation actions, so core DevOps workflows like applying Terraform changes or upgrading Helm releases cannot be completed.

    Maintenance

    ActivityNo data
    ResponsivenessSyncing

    Related MCP Connectors

    Related MCP Servers