gpt-vis-mcp by yaonyan - MCP Server

GPT-Vis MCP Server

A local/private-favored wrapper for antvis/mcp-server-chart that generates charts locally without external server dependencies.

✨ Features

🔒 Private & Secure: Local chart generation, no external dependencies
🚀 Easy Setup: One command installation for Claude Desktop
🎨 Rich Charts: 20+ chart types (pie, line, bar, radar, maps, etc.)
📊 Enterprise Ready: Perfect for secure environments

🚀 Usage

Claude Desktop (Recommended)

Add to your Claude Desktop MCP settings:

Option 1: NPX (Recommended)

{
  "mcpServers": {
    "gpt-vis-mcp": {
      "command": "npx",
      "args": ["-y", "@jsr2npm/yao__gpt-vis-mcp@latest"]
    }
  }
}

Option 2: Deno Direct

{
  "mcpServers": {
    "gpt-vis-mcp": {
      "command": "deno",
      "args": ["run", "--allow-all", "@yao/gpt-vis-mcp/bin"]
    }
  }
}

Option 3: NPX + Deno

{
  "mcpServers": {
    "gpt-vis-mcp": {
      "command": "npx",
      "args": ["-y", "deno", "run", "--allow-all", "@yao/gpt-vis-mcp/bin"]
    }
  }
}

You may experience canvas dependency issues and font rendering issues when using npx. If so, try option 4.

Option 4: Docker

{
  "mcpServers": {
    "gpt-vis-mcp": {
      "command": "docker",
      "args": [
        "run",
        "--interactive",
        "--rm",
        "-v",
        "/tmp/gpt-vis-charts:/tmp/gpt-vis-charts",
        "ghcr.io/yaonyan/gpt-vis-mcp:latest-mcp"
      ]
    }
  }
}

Direct Command Line Usage

You can also run the server directly with various transport modes:

# Using JSR package (recommended)
# Default stdio mode (for MCP clients)
deno run --allow-all @yao/gpt-vis-mcp/bin

# SSE server mode
deno run --allow-all @yao/gpt-vis-mcp/bin --transport sse --port 3000 --host localhost

# Short form
deno run --allow-all @yao/gpt-vis-mcp/bin -t sse -p 3000 -h localhost

# Show help
deno run --allow-all @yao/gpt-vis-mcp/bin --help

# Alternative: Using npx with deno
# Default stdio mode
npx -y deno run --allow-all @yao/gpt-vis-mcp/bin

# SSE server mode  
npx -y deno run --allow-all @yao/gpt-vis-mcp/bin --transport sse --port 3000

# Show help
npx -y deno run --allow-all @yao/gpt-vis-mcp/bin --help

Transport Modes:

stdio: Direct MCP protocol communication via stdin/stdout (default)
sse: Server-Sent Events HTTP server for web clients

Docker with SSE Mode

# Run with SSE server
docker run -p 3000:3000 ghcr.io/yaonyan/gpt-vis-mcp:latest-mcp --transport sse --port 3000 --host 0.0.0.0

# Access the SSE endpoint
curl http://localhost:3000/sse

Set environment variables as needed:

Variable	Description	Default
`RENDERED_IMAGE_PATH`	Chart images directory	system temp
`RENDERED_IMAGE_HOST_PATH`	Base URL for accessing images	(optional)

Docker SSR Server

We also provide an SSR Server that follows the requirements of https://github.com/antvis/mcp-server-chart?tab=readme-ov-file#-private-deployment

# Run SSR API server
docker run -p 3000:3000 -e RENDERED_IMAGE_HOST_PATH=http://localhost:3000/charts ghcr.io/yaonyan/gpt-vis-mcp:latest-http

# Test the SSR API
❯ curl -X POST http://localhost:3000/generate \
             -H "Content-Type: application/json" \
             -d '{"type": "pie", "data": [{"category": "A", "value": 30}, {"category": "B", "value": 70}]}'
{"success":true,"resultObj":"http://localhost:3000/charts/chart_1750500506056_T6IC0Vtp.png"}

You can then use the SSR server with the upstream @antv/mcp-server-chart by specifying the VIS_REQUEST_SERVER environment variable:

{
  "mcpServers": {
    "mcp-server-chart": {
      "command": "npx",
      "args": ["-y", "@antv/mcp-server-chart"],
      "env": {
        "VIS_REQUEST_SERVER": "http://localhost:3000/generate"
      }
    }
  }
}

This allows you to use the original MCP server while leveraging your private SSR endpoint for chart generation.

VIS_REQUEST_SERVER example

🤝 Contributing

See for development setup.

📄 License

MIT License - see