Skip to content

MCP Server for AI Agents

cloud-audit includes a Model Context Protocol (MCP) server that lets AI coding assistants scan your AWS account directly.

Quick Setup

Claude Code

claude mcp add cloud-audit -- uvx --from cloud-audit cloud-audit-mcp

Project-scoped config

Add .mcp.json to your repo root (shared with team):

{
  "cloud-audit": {
    "command": "uvx",
    "args": ["--from", "cloud-audit", "cloud-audit-mcp"]
  }
}

With pip

pip install cloud-audit
claude mcp add cloud-audit -- cloud-audit-mcp

Available Tools

Tool Description
scan_aws Run a full security scan
get_findings Get findings filtered by severity/category
get_attack_chains Get detected attack chains
get_remediation Get fix for a specific check ID
get_health_score Get the health score and risk exposure
list_checks List all available checks

Example Prompts

After setup, ask your AI assistant:

  • "Scan my AWS account and show me the critical findings"
  • "What attack chains were detected?"
  • "How much risk exposure does my account have in dollars?"
  • "Show me the Terraform code to fix aws-iam-001"
  • "Which CIS controls are we failing?"

How It Differs

Prowler and Wiz have MCP servers, but both require their paid SaaS platform ($99+/month). cloud-audit MCP works locally with zero accounts, zero API keys, and zero data sent anywhere.

Compatibility

Works with any MCP-compatible client:

  • Claude Code
  • Cursor
  • VS Code with GitHub Copilot
  • Any tool supporting the Model Context Protocol

Requirements

  • AWS credentials configured (same as CLI usage)
  • SecurityAudit IAM policy attached to your role/user