Back to .md Directory

Configure HoloViz MCP for Windsurf

Walks through adding HoloViz MCP as a server in Windsurf IDE's JSON config file and testing it with AI queries.

May 2, 2026
0 downloads
2 views
ai mcp
View source

What this file does

Walks through adding HoloViz MCP as a server in Windsurf IDE's JSON config file and testing it with AI queries.

When to use it

  • You use Windsurf IDE and want Panel/HoloViz AI assistance
  • You need to configure an MCP server in Windsurf's config.json
  • You are debugging why Windsurf AI doesn't recognize HoloViz components

Assumes this stack

Windsurf IDEHoloViz MCPPanelJSON

Configure HoloViz MCP for Windsurf

This guide shows you how to configure HoloViz MCP with Windsurf IDE.

Prerequisites

Locate Your Configuration File

Find your Windsurf MCP configuration file. The location varies by operating system and Windsurf version. Common locations include:

  • macOS: ~/Library/Application Support/Windsurf/config.json or similar
  • Windows: %APPDATA%\Windsurf\config.json or similar
  • Linux: ~/.config/Windsurf/config.json or similar

!!! note The exact configuration file location may vary. Check Windsurf's documentation for your specific version.

Add the MCP Server

  1. Open the Windsurf MCP configuration file in a text editor
  2. Add this configuration:
{
  "mcpServers": {
    "holoviz": {
      "command": "holoviz-mcp"
    }
  }
}
  1. Save the file
  2. Restart Windsurf

!!! tip "Existing Configuration" If your file already has other MCP servers configured, add "holoviz" to the existing mcpServers object:

```json
{
  "mcpServers": {
    "existing-server": {
      "command": "existing-command"
    },
    "holoviz": {
      "command": "holoviz-mcp"
    }
  }
}
```

Test Your Configuration

Test the configuration with Windsurf's AI:

  1. Simple Query - Open Windsurf's AI assistant and ask:

    List available Panel input components
    
  2. Detailed Query - Ask for specific information:

    What parameters does the Panel TextInput component have?
    
  3. Code Generation - Ask Windsurf AI to generate code:

    Create a simple Panel dashboard with a slider
    

If Windsurf's AI provides detailed, accurate responses with specific Panel component information, your configuration is working! 🎉

Advanced Configuration

Set Log Level

For debugging, increase the log level:

{
  "mcpServers": {
    "holoviz": {
      "command": "holoviz-mcp",
      "env": {
        "HOLOVIZ_MCP_LOG_LEVEL": "DEBUG"
      }
    }
  }
}

Custom Configuration Directory

Use a custom directory for configuration and data:

{
  "mcpServers": {
    "holoviz": {
      "command": "holoviz-mcp",
      "env": {
        "HOLOVIZ_MCP_USER_DIR": "/path/to/custom/dir"
      }
    }
  }
}

Troubleshooting

Configuration File Not Found

Check Windsurf's documentation for the correct configuration file location for your version.

Create the file if it doesn't exist, using the configuration above.

Server Not Starting

Check that holoviz-mcp is installed:

holoviz-mcp --version

Test the server directly:

holoviz-mcp

Press Ctrl+C to stop it.

Windsurf AI Not Recognizing Components

  1. Verify documentation index exists:

    ls ~/.holoviz-mcp
    
  2. Recreate the index:

    holoviz-mcp update index
    
  3. Restart Windsurf

Configuration Not Loading

Verify JSON syntax - ensure there are no trailing commas, missing quotes, or other syntax errors.

Restart Windsurf completely - quit and reopen the application.

Check logs - look for error messages in Windsurf's output or logs.

Permission Errors

Linux/macOS: Ensure the configuration file is readable:

chmod 644 ~/.config/Windsurf/config.json  # Adjust path as needed

Next Steps

What's inside

5 sections: prerequisites, config location, server addition, testing, troubleshooting, plus advanced options and next steps

Change this for your project

  • Replace holoviz-mcp with your own MCP server command if not using HoloViz
  • Replace "holoviz" key in JSON with your server's name
  • Replace ~/.holoviz-mcp path in troubleshooting commands with your server's data directory

Where it goes

Save in docs/ or the repository root. Gives agents and new contributors a map of the codebase.

Worth borrowing

  • Testing MCP with three escalating query types: simple, detailed, code generation
  • Offering both fresh config and merge-into-existing config examples

Related Documents