kloudhub-mcp 0.5.0

MCP server for KloudHub - Control your cloud infrastructure with AI.

kloudhub-mcp

MCP server for KloudHub. Connect Claude, Codex, or any MCP-compatible AI client to KloudHub and manage your cloud infrastructure through natural language.

What you can do

Ask Claude...	What happens
"Scan my AWS account and show me what's wasting money"	Discovers all resources across regions, runs cost and waste analysis
"Generate a VPC for my new service and deploy it"	Creates a Terraform blueprint and deploys it to AWS
"Show me all failed deployments this week"	Queries deployment history with status and logs
"Run KloudPilot on my account and give me the top recommendations"	Starts an AI health analysis with per-service breakdown
"What's my cloud spend vs last month?"	Returns MTD costs and month-over-month comparison
"Drill into my EC2 findings and dismiss the accepted risks"	Per-service analysis with dismiss and snooze workflows
"Start a cloud shell linked to my AWS account"	Provisions a shell with AWS credentials pre-loaded
"Run a team analysis across all accounts in my catalog"	Aggregated KloudPilot analysis for catalogs

Tools (45+)

• Cloud Accounts - list connected AWS accounts
• Klouds - list, inspect, and generate infrastructure blueprints
• Deployments - deploy, cancel, destroy, get status and logs
• Cloud Scanning - discover all AWS resources across regions
• Cost Analysis - MTD costs, forecasts, trends, month-over-month comparison
• Waste Detection - identify idle and oversized resources with savings estimates
• KloudPilot - AI health scores, per-service drill-down, dismiss/snooze findings, team analysis
• Persistent Shells - create cloud-linked or sandbox shells, share links, read files

Quickstart

1. Generate an API key

Go to kloudhub.io/mcp and generate an API key.

2. Add to Claude Code

{
  "mcpServers": {
    "kloudhub": {
      "type": "stdio",
      "command": "uvx",
      "args": ["kloudhub-mcp"],
      "env": {
        "KLOUDHUB_API_KEY": "khub_live_YOUR_KEY"
      }
    }
  }
}

Add to ~/.claude/settings.json (global) or .claude/settings.json (per-project).

3. Add to Claude Desktop

{
  "mcpServers": {
    "kloudhub": {
      "command": "uvx",
      "args": ["kloudhub-mcp"],
      "env": {
        "KLOUDHUB_API_KEY": "khub_live_YOUR_KEY"
      }
    }
  }
}

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

Restart Claude Desktop after saving.

4. Claude Web / Remote MCP (OAuth)

In Claude.ai (or any MCP client that supports remote OAuth), go to Settings, Integrations, Add MCP Server and enter:

https://mcp.kloudhub.io/mcp

You will be redirected to KloudHub to log in. After signing in, the connection is established automatically. No API key needed. An API key named "Claude.ai connector" will appear in your KloudHub settings so you can revoke it to disconnect at any time.

This works with any MCP client that supports OAuth 2.0 + PKCE.

5. Codex CLI

{
  "mcpServers": {
    "kloudhub": {
      "command": "uvx",
      "args": ["kloudhub-mcp"],
      "env": {
        "KLOUDHUB_API_KEY": "khub_live_YOUR_KEY"
      }
    }
  }
}

Add to ~/.codex/config.json.

Security

• Cloud credentials (AWS keys, secrets, tokens) are never accessible via MCP
• Only operational data is returned: resource names, costs, analysis results, deployment logs
• API keys are stored as SHA-256 hashes and shown only once at creation
• Remote connections use OAuth 2.0 with PKCE (no secrets in URLs or configs)
• All tool calls go through KloudHub's authenticated API with per-request authorization

Requirements

• Python 3.11+
• A KloudHub account with at least one connected AWS cloud account

Links

• KloudHub
• Generate API Key
• Documentation