CloudQuery MCP

CloudQuery MCP Server lets AI assistants query your CloudQuery asset inventory data, ask questions about your cloud infrastructure, find savings or security issues.

GitHub

Dokumentation

CloudQuery MCP Server

A Model Context Protocol (MCP) server that provides access to CloudQuery asset inventory data.

CloudQuery Platform customers: connect to the hosted server

You do not need this binary to use CloudQuery Platform. Your platform serves the MCP endpoint itself at https://<your-platform-url>/mcp. An admin enables it once under Settings → Organization → MCP server, and that page shows the snippets for your deployment.

There are two ways to connect:

Sign in with your CloudQuery account (recommended). For clients that can open a browser, such as Claude Code, Claude Desktop and ChatGPT. You sign in and approve access once, and the client keeps a token that refreshes itself. You can approve read-only access.
```
claude mcp add --transport http cloudquery https://<your-platform-url>/mcp
```
```
{
  "mcpServers": {
    "cloudquery": {
      "type": "http",
      "url": "https://<your-platform-url>/mcp"
    }
  }
}
```

Send an API key as a static header. For clients that cannot open a browser, such as Cursor or CI jobs.

claude mcp add --transport http cloudquery https://<your-platform-url>/mcp \
  --header "Authorization: Bearer <YOUR_API_KEY>"

{
  "mcpServers": {
    "cloudquery": {
      "type": "http",
      "url": "https://<your-platform-url>/mcp",
      "headers": { "Authorization": "Bearer <YOUR_API_KEY>" }
    }
  }
}

Full instructions: https://www.cloudquery.io/docs/platform/features/mcp-server.

Use the binary below if you are a CLI user, or if your client can only start a local process.

Prerequisites

Claude Desktop, Cursor IDE or any other tool capable of using MCP servers.
If you're using it against CloudQuery Platform, a non-expired CloudQuery Platform API Key and the URL of the deployment it is valid for.
If you're using it against a PostgreSQL destination (from one or more CloudQuery syncs), the PostgreSQL connection string.

Installation

From MCP Registry (Recommended)

The CloudQuery MCP server is available through the official MCP Registry. This is the recommended way to install and use the server.

For Claude Desktop users:

Open Claude Desktop
Go to Settings → MCP Servers
Add a new server with the identifier: io.github.cloudquery/mcp
Configure the appropriate environment variables for your use case (see configuration sections below)

For other MCP clients: Use the server identifier io.github.cloudquery/mcp to install from the MCP Registry.

Download Binary

Download the latest binary for your platform from the Releases page.

macOS Security Note

On macOS, you may encounter a security warning when running the binary for the first time. This is because the binary is not signed with an Apple Developer certificate. To bypass this:

Option 1 (Recommended): Use the command line to remove the quarantine attribute:

xattr -d com.apple.quarantine /path/to/cq-platform-mcp

Option 2: Go to System Preferences > Security & Privacy > General, and click "Allow Anyway" for the blocked app.

Supported modes

The MCP server supports four modes:

cli - for getting started with CloudQuery CLI and generate configuration files using natural language
postgres - for CLI users syncing to a PostgreSQL destination, to query the data synced to the database using natural language
snowflake - for CLI users syncing to a Snowflake destination, to query the data synced to the database using natural language
platform - for CloudQuery Platform customers to query the data synced to the platform using natural language

CLI Mode

If no environment variables are set, the MCP server will default to CLI mode.

PostgreSQL Mode

Configure the following environment variables to enable PostgreSQL mode.

POSTGRES_CONNECTION_STRING - postgres://user:password@host:port/database

Note: If no search_path is specified in the connection string, the MCP server automatically defaults to public schema. You can override this by explicitly setting search_path in your connection string (e.g., postgres://user:password@host:port/database?search_path=myschema or host=localhost dbname=mydb search_path=myschema).

Kerberos Authentication

For Kerberos/GSSAPI authentication, include the appropriate parameters in the connection string:

postgres://username@REALM.EXAMPLE.COM@host:port/database?krbsrvname=postgres
postgres://username@host:port/database?krbsrvname=postgres&gsslib=gssapi

The MCP server supports reading .env files.

Snowflake Mode

Configure the following environment variables to enable Snowflake mode.

SNOWFLAKE_CONNECTION_STRING - Snowflake connection string in the format: user:password@account/database/schema?warehouse=warehouse_name

Example connection strings:

user:password@account/database/schema - Basic authentication
user:password@account/database/schema?warehouse=COMPUTE_WH - With warehouse specified
user:password@account.region/database/schema?warehouse=COMPUTE_WH - With region

For more connection string options, see the Snowflake Go Driver documentation.

The MCP server supports reading .env files.

Platform Mode

Configure the following environment variables to enable Platform mode.

CQ_PLATFORM_API_URL - https://your-deployment.cloudquery.io/api
CQ_PLATFORM_API_KEY - Your CloudQuery Platform API key

When the tenant has the hosted MCP server enabled (Settings -> Organization -> MCP server), the binary proxies <CQ_PLATFORM_API_URL>/mcp and exposes its full tool catalog; clients that support remote servers can connect to that URL directly with the same API key. Otherwise the binary serves its built-in tools.

Built-in tool names and their hosted equivalents:

Built-in	Hosted
`execute-clickhouse-sql-query`	`run_query`
`table-search-regex`	`search_tables`
`table-schemas`	`get_table_schema`
`known-good-queries`	`list_example_queries`
`list-installed-plugins`	`describe_inventory`

The MCP server supports reading .env files.

Optional Environment Variables

CQAPI_LOG_LEVEL - Log level: debug, info, warn, error (default: info)
LOG_PRETTY - Enable pretty console logging (default: auto-detected based on TTY)
LOG_COLOR - Enable colored logging (default: auto-detected based on TTY)

Claude Desktop Integration

Configure Claude Desktop by editing the config file at $HOME/Library/Application\ Support/Claude/claude_desktop_config.json on MacOS or $HOME/.config/Claude/claude_desktop_config.json on Linux:

If you're using it against CloudQuery Platform:

{
  "mcpServers": {
    "cloudquery": {
      "command": "/absolute/path/to/cq-platform-mcp",
      "args": [],
      "env": {
        "CQ_PLATFORM_API_KEY": "your_api_key_here",
        "CQ_PLATFORM_API_URL": "https://your-instance.cloudquery.io/api"
      }
    }
  }
}

If you're using it against a PostgreSQL destination, e.g.:

{
  "mcpServers": {
    "cloudquery": {
      "command": "/absolute/path/to/cq-platform-mcp",
      "args": [],
      "env": {
        "POSTGRES_CONNECTION_STRING": "postgres://user:password@localhost:5432/database?sslmode=disable"
      }
    }
  }
}

If you're using it against a Snowflake destination, e.g.:

{
  "mcpServers": {
    "cloudquery": {
      "command": "/absolute/path/to/cq-platform-mcp",
      "args": [],
      "env": {
        "SNOWFLAKE_CONNECTION_STRING": "user:password@account/database/schema?warehouse=COMPUTE_WH"
      }
    }
  }
}

Notes:

"command" must be an absolute path to your local cq-platform-mcp binary.

Example: "/home/username/cq-platform-mcp" or "/Users/username/cq-platform-mcp"

Cursor IDE Integration

Configure Cursor IDE by adding the MCP server configuration to your settings:

Open Cursor IDE
Go to Settings > Cursor Settings > Tools and Integrations
Add a new MCP server with the following configuration:

{
  "mcpServers": {
    "cloudquery": {
      "command": "/path/to/mcp/binary",
      "args": [],
      "env": {
        "CQ_PLATFORM_API_KEY": "your_api_key_here",
        "CQ_PLATFORM_API_URL": "https://your-instance.cloudquery.io/api"
      }
    }
  }
}

If you're using it against a PostgreSQL destination, e.g.:

{
  "mcpServers": {
    "cloudquery": {
      "command": "/path/to/mcp/binary",
      "args": [],
      "env": {
        "POSTGRES_CONNECTION_STRING": "postgres://user:password@localhost:5432/database?sslmode=disable"
      }
    }
  }
}

If you're using it against a Snowflake destination, e.g.:

{
  "mcpServers": {
    "cloudquery": {
      "command": "/path/to/mcp/binary",
      "args": [],
      "env": {
        "SNOWFLAKE_CONNECTION_STRING": "user:password@account/database/schema?warehouse=COMPUTE_WH"
      }
    }
  }
}

VSCode IDE Integration

There are multiple ways to integrate MCP servers in VSCode, as described in the VSCode MCP documentation. A summary of the steps to integrate the CloudQuery MCP server in VSCode is:

Open VSCode IDE
Go to MCP: Open User Configuration (can also be done at the workspace or folder level) via the keyboard shortcut Ctrl+Shift+P (Windows/Linux) or Cmd+Shift+P (Mac)
Add a new MCP server with the following configuration:

If you're using it against CloudQuery Platform:

{
  "inputs":[
    {
      "type":"promptString",
      "id":"cloudquery-platform-api-key",
      "description":"CloudQuery Platform API Key",
      "password":true
    },
    {
      "type":"promptString",
      "id":"cloudquery-platform-api-url",
      "description":"CloudQuery Platform API URL",
      "password":false
    }
  ],
  "servers":{
    "CloudQuery":{
      "type":"stdio",
      "command":"/path/to/mcp/binary",
      "env":{
        "CQ_PLATFORM_API_KEY": "${input:cloudquery-platform-api-key}",
        "CQ_PLATFORM_API_URL": "${input:cloudquery-platform-api-url}"
      }
    }
  }
}

If you're using it against a PostgreSQL destination, e.g.:

{
  "inputs":[
    {
      "type":"promptString",
      "id":"cloudquery-postgres-connection-string",
      "description":"CloudQuery Postgres Connection String",
      "password":true
    }
  ],
  "servers":{
    "CloudQuery":{
      "type":"stdio",
      "command":"/path/to/mcp/binary",
      "env":{
        "POSTGRES_CONNECTION_STRING": "${input:cloudquery-postgres-connection-string}"
      }
    }
  }
}

If you're using it against a Snowflake destination, e.g.:

{
  "inputs":[
    {
      "type":"promptString",
      "id":"cloudquery-snowflake-connection-string",
      "description":"CloudQuery Snowflake Connection String",
      "password":true
    }
  ],
  "servers":{
    "CloudQuery":{
      "type":"stdio",
      "command":"/path/to/mcp/binary",
      "env":{
        "SNOWFLAKE_CONNECTION_STRING": "${input:cloudquery-snowflake-connection-string}"
      }
    }
  }
}

To see the new MCP server in the list, you might need to restart VSCode.

On first use, you will be prompted to enter the API key and API URL, or the PostgreSQL connection string.

Streamable HTTP Server

The MCP server can also be run as a Streamable HTTP server, which allows you to connect to it remotely. To enable this, set the HTTP_ADDRESS environment variable to the desired address and port, e.g.:

export HTTP_ADDRESS=":8080"

The server will then listen for incoming HTTP requests under the path /mcp on the specified address and port.