@alpacacloud/mcp

MCP (Model Context Protocol) server for Alpaca Cloud — lets AI agents like Claude, Cursor, and Windsurf connect directly to your project database, auth, and storage.

Quick Start

Claude Desktop

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

{
  "mcpServers": {
    "alpaca-cloud": {
      "command": "npx",
      "args": [
        "@alpacacloud/mcp",
        "--url=https://YOUR-PROJECT.alpaca-cloud.com",
        "--key=YOUR-SERVICE-ROLE-KEY"
      ]
    }
  }
}

Cursor / Windsurf

Add to your .cursor/mcp.json or .windsurf/mcp.json:

{
  "mcpServers": {
    "alpaca-cloud": {
      "command": "npx",
      "args": ["@alpacacloud/mcp", "--url=https://YOUR-PROJECT.alpaca-cloud.com", "--key=YOUR-SERVICE-ROLE-KEY"]
    }
  }
}

Environment Variables

export ALPACA_URL=https://your-project.alpaca-cloud.com
export ALPACA_KEY=your-service-role-key
npx @alpacacloud/mcp

Available Tools

| Tool | Description | |------|-------------| | run_sql | Execute any SQL (SELECT, INSERT, DDL, etc.) | | list_tables | List all tables with columns | | get_table | Full schema for a specific table | | query_table | Query with filters, ordering, limits | | insert_row | Insert one or many rows | | update_rows | Update rows matching a filter | | delete_rows | Delete rows matching a filter | | list_auth_users | List all auth users | | get_project_info | Get endpoints and API keys | | list_storage_buckets | List storage buckets | | list_storage_objects | List files in a bucket |

Get Your Keys

1. Go to alpaca-cloud.com and open your project
2. Click Connect → Easy Connect
3. Copy your Project URL and Service Role Key

What Can You Do?

Ask Claude or Cursor things like:

• "Show me all tables in my database"
• "Create a users table with email, name, and created_at"
• "Insert 5 test rows into the products table"
• "What are the last 10 orders from the orders table?"
• "List all auth users who signed up this week"

Security

• Your service role key has full database access — keep it secret
• The MCP server runs locally on your machine and communicates directly with your project
• No data passes through any third-party servers