Overview

Managing infrastructure through an API is powerful, but it requires knowing the right endpoints, structuring the right payloads, and writing the right scripts. For most requests, that means significant time spent before a single resource is touched.

The Scalr MCP Server removes that friction by connecting your AI assistant directly to the Scalr API. Instead of compiling API calls manually, you describe what you need in plain language, and the MCP server handles the translation. Requests that would have taken hours or days of scripting now take minutes.

The server runs remotely or locally as a containerized service and works with any MCP-compatible client. The documentation below includes configuration instructions for specific LLM clients like Claude Desktop and VSCode.

Remote MCP Server

The Scalr remote MCP server is the simplest way to get started. It requires no Docker, no local setup, and no infrastructure to maintain. Just enable it in your account settings and connect your AI client.

Enable the Integration

• Go to Account Settings → Integrations → AI Assistant
• Toggle on Enable Scalr MCP

Connect Your AI Client

Point your MCP client at https://scalr.io/mcp. Authentication is handled via OAuth 2.1, and the client will guide you through login and consent. Once connected, your AI agent has access to all Scalr tools scoped to your account and your user's permissions.

Claude Example

In Claude, go to Settings, Connectors, and click "Add custom connector". Add https://scalr.io/mcp as the URL for the MCP server. Click connect, and you will be redirected to Scalr to authenticate.

Cursor Example

In Cursor, go to settings, "Tools & MCP", click New MCP server. When the editor opens, add the following in the mcp.json:

{
  "mcpServers": {
    "scalr-remote": {
      "url": "https://scalr.io/mcp",
      "headers": {
        "Content-Type": "application/json"
      }
   }
  }
 }

Save the file, go back to settings, click connect, and you will be redirected to Scalr to authenticate.

Local MCP Server

HTTP Configuration

The HTTP configuration runs wherever the Docker image is pulled and installed. Users can connect to it over a network without requiring local installation on their workstations.

Prerequisites

Please review the following prerequisites before configuring the MCP server.

• Docker installed and running wherever the MCP server will run. Docker must be running when starting the MCP server.
• Scalr API token (Personal Access Token or Service Account Token). Learn how to generate a token here. It is recommended to create a dedicated API token for the MCP server use with the minimum required permissions.
• MCP-compatible client (e.g., Claude Desktop) to connect to the MCP server.

Configuring the MCP Server

To install and configure the MCP server, the following Docker command should be executed wherever the MCP server will run:

docker run -d   
--pull always   
--name scalr-mcp-server   
-p 8000:8000   
-e SCALR_API_URL=[https://your-account.scalr.io](https://your-account.scalr.io)   
-e MCP_TRANSPORT=http   
-e MCP_HTTP_HOST=0.0.0.0   
-e MCP_LOG_FILE=/var/log/mcp/server.log   
-v ~/mcp-logs:/var/log/mcp   
scalr/mcp-server:latest

Replace your-account.scalr.io with your actual Scalr.io URL

Authentication:

Authentication for the MCP server can be added as a variable in the Docker command above or set up per request when adding the connection to your LLM client (See below).

• Clients provide a token via the Authorization: Bearer <token> header (per-request in the LLM client)
• Or set SCALR_API_TOKEN environment variable as global fallback
• Per-request tokens take precedence over global tokens

Security:

• Configure CORS: -e MCP_CORS_ORIGINS='[https://your-app.com](https://your-app.com)
• Restrict hosts: -e MCP_HTTP_ALLOWED_HOSTS='your-domain.com'

Example LLM Configurations

Cursor IDE (HTTP)

Open your MCP JSON file for Cursor and update it as seen below.

File location:

• Mac/Linux: ~/.cursor/config.json
• Windows: %APPDATA%\Cursor\config.json

{
  "mcpServers": {
    "scalr-local": {
      "url": "http://localhost:8000/mcp",
      "headers": {
        "Authorization": "Bearer <your token>",
        "Content-Type": "application/json"
      }
    }
  }
}

#

Make sure to replace the url value with your actual URL and the "Bearer <your token>" with your token.

In Cursor, you can validate that the MCP server is running and execute a test prompt to make sure you are receiving answers back:

STDIO Configuration

The STDIO configuration runs as local subprocesses on the machine where it is configured.

Prerequisites

Regardless of the MCP client you are using, you will need to review the following prerequisites before using the MCP server.

• Docker installed and running on your machine. Docker must be running when starting the MCP server.
• Scalr API token (Personal Access Token or Service Account Token). Learn how to generate a token here. It is recommended to create a dedicated API token for the MCP server use with the minimum required permissions.
• MCP-compatible client (e.g., Claude Desktop)

Example LLM Configurations

Choose between Claude, Cursor, or VS Code depending on what you have installed locally:

The following section provides information specific to configuring the Claude desktop app and running the MCP server locally. To start, add the following configuration to your Claude Desktop config file:

1. Open Claude Settings
2. Navigate to "Developer"
3. Click "Add MCP Server"
4. Add the following to the config file.

JSON

{
  "mcpServers": {
    "scalr": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "--pull=always",
        "--env",
        "SCALR_API_TOKEN=your_api_token_here",
        "--env",
        "SCALR_API_URL=https://your-account.scalr.io",
        "scalr/mcp-server:latest"
      ]
    }
  }
}

You must replace the following values:

• your_api_token_here with your actual Scalr API token
• your-account.scalr.io with your Scalr instance URL (e.g., mycompany.scalr.io)

Before restarting Claude, ensure that Docker is running on your machine. Once confirmed, close and reopen Claude Desktop to load the new configuration. Start with a simple question like "List all of my Scalr environments" to make sure that it is pulling the information back from the Scalr API:

Available MCP Tools

MCP tools give large language models (LLMs) the ability to perform actions by calling executable functions exposed by servers. The following tools are available for the Scalr MCP server.

Note: All endpoints marked with x-mcp-tool: true in the Scalr Public OpenAPI Specification are available as MCP tools.

Troubleshooting

Claude Desktop doesn't show Scalr tools

• Check configuration file syntax: Ensure your JSON is valid (use a JSON validator)
• Restart Claude Desktop: Completely quit and reopen the application
• Check Docker: Ensure Docker is running: docker ps
• View logs: Check Claude Desktop logs for errors:
  • macOS: ~/Library/Logs/Claude/mcp-server-scalr.log
  • Windows: %APPDATA%\Claude\Logs\mcp-server-scalr.log
  • Linux: ~/.config/Claude/logs/mcp-server-scalr.log

Authentication Errors

• Verify API token: Test your token with curl:

curl -H "Authorization: Bearer YOUR_TOKEN" \
     https://your-account.scalr.io/api/iacp/v3/account

• Check token permissions: Ensure the token has appropriate read/write permissions

Docker pull errors

If you see "pull access denied" or network errors:

# Manually pull the image to test connectivity
docker pull scalr/mcp-server:latest

# Check if image exists
docker images | grep scalr/mcp-server

Tools not responding

• Check API connectivity: Ensure you can reach your Scalr instance
• Review rate limits: Check if you're hitting API rate limits
• Increase timeout: Docker might need more time on slower connections

Example Prompts

Here are some example questions you can ask Claude:

Discovery & Exploration

• "Show me all workspaces in the production environment."
• "Which Terraform modules are most used?"
• "What versions of Terraform are running across my infrastructure?"

Security

• "Review token usage in my account. Show tokens that don't have owners, have never been rotated, and don't have a description. Separate it into three outputs."
• "Suggest ways to tighten my Scalr account security"
• "Pull all users in my Scalr account that have an admin access policy."
• "How many tokens have not been rotated in the last 30 day?s"

Monitoring & Insights

• "Show me recent runs that failed."
• "What policy checks have failed recently?"
• "Show me usage statistics for AWS provider."

Management

• "Create a workspace named 'new-app-staging' in the dev environment"
• "Show me details about workspace ws-abc123"
• "List all VCS providers configured in my account"

Analytics

• "Which Terraform resources are most commonly used?"
• "Show me module usage by namespace"
• "What provider versions are in use?"
• "List all access tokens and their usage"

Support & Resources

• Scalr API Documentation: https://docs.scalr.io/reference/overview-1
• Model Context Protocol: https://modelcontextprotocol.io/
• Docker Hub: https://hub.docker.com/r/scalr/mcp-server

Demo