Setting Up Flow AI Gateway with Claude Desktop

The following step-by-step guide will help you to get Claude desktop connected to the Flow AI gateway

Prerequisites

1. Claude Desktop: Ensure you have the latest version installed (Node js will also need to be installed)
2. Flow Software: Running Flow AI Gateway MCP server. The Flow AI Gateway is bundled within the Flow Server. Thus, if the Flow Server is deployed, the Flow AI gateway will be running.
3. Licensing: the Flow AI Gateway is a licensed component. Ensure that is licensed for your Flow instance. A Demo Flow system will enable the Flow AI Gateway for 14 days
4. Network Access: Connectivity between Claude Desktop and Flow AI Gateway

Want to follow along with an actual build out? Check the YouTube video below:

Configure Claude Desktop Config File

Manual Configuration

Step 1: Locate Configuration File

The configuration file location depends on your operating system:

macOS:

~/Library/Application Support/Claude/claude_desktop_config.json

Windows:

%APPDATA%\Claude\claude_desktop_config.json

Step 2: Add Flow AI Gateway Server Configuration

Edit the claude_desktop_config.json file to include your Flow AI Gateway:

{
  "mcpServers": {
    "Flow_MCP": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "http://<YourFlowServerIPaddressHere>:<YourFlowServerPortHere>/mcp/sse",
        "--transport",
        "sse-only",
        "--allow-http"
      ],
      "timeout": 30000,
      "initTimeout": 10000
    }
  }
}

Configuration Parameters Explained:

• command: "npx": Uses npx to run the mcp-remote package
• "-y": Automatically confirms package installation if not present
• "mcp-remote": The remote MCP client package
• "http://<YourFlowServerIPaddressHere>:<YourFlowServerPortHere>/mcp/sse": Your Flow AI Gateway server URL with SSE endpoint
• "--transport", "sse-only": Uses Server-Sent Events for real-time communication
• "--allow-http": Permits HTTP connections (use HTTPS in production)
• "timeout": 30000: Connection timeout in milliseconds (30 seconds)
• "initTimeout": 10000: Initialization timeout in milliseconds (10 seconds)

Step 3: Customize for Your Environment

Replace <YourFlowServerIPaddressHere>:<YourFlowServerPortHere> with your actual Flow server hostname and port:

Step 4: Restart Claude Desktop

After saving the configuration file, completely restart Claude Desktop to load the new MCP server connection.

Troubleshooting Setup Issues

Tools Not Available After Installation

Restart Claude Desktop to refresh the extension registry.

• Restart Claude Desktop completely
• Verify Flow AI Gateway server is running

• Check network connectivity to Flow server

   

Permission Errors

On macOS, check System Preferences > Security & Privacy if you receive security warnings. On Windows, ensure Claude Desktop has necessary permissions to access required directories.

Using Flow AI Gateway

Basic Workflow

1. Discovery Phase: Search for equipment, metrics, and events
2. Planning Phase: Use get_execution_plan for complex analyses
3. Data Retrieval: Access historical and real-time data
4. Analysis Phase: Interpret results with contextual business understanding

Example Use Cases

Production Analysis:

"Show me the hourly production totals for Filler 01 over the last week"

Equipment Performance:

"What was the efficiency of all packaging lines during the morning shift yesterday?"

Quality Monitoring:

"Compare the defect rates across all production lines for the current month"

Maintenance Planning:

"Which equipment had the most downtime events in the past quarter?"

Best Practices

1. Be Specific: Use exact equipment names and time ranges
2. Think Hierarchically: Understand the folder → metric → measure structure
3. Use UTC Timestamps: Always specify time zones clearly
4. Leverage Pre-calculated Data: Prefer measure data over raw tag aggregation