Mermaid MCP Server

A Model Context Protocol (MCP) server that converts Mermaid diagrams to PNG images or SVG files. This server allows AI assistants and other applications to generate visual diagrams from textual descriptions using the Mermaid markdown syntax.

Features

• Converts Mermaid diagram code to PNG images or SVG files
• Supports multiple diagram themes (default, forest, dark, neutral)
• Customizable background colors
• Uses Puppeteer for high-quality headless browser rendering
• Implements the MCP protocol for seamless integration with AI assistants
• Flexible output options: return images/SVG directly or save to disk
• Error handling with detailed error messages

How It Works

The server uses Puppeteer to launch a headless browser, render the Mermaid diagram to SVG, and optionally capture a screenshot of the rendered diagram. The process involves:

1. Launching a headless browser instance

2. Creating an HTML template with the Mermaid code

3. Loading the Mermaid.js library

4. Rendering the diagram to SVG

5. Either saving the SVG directly or taking a screenshot as PNG

6. Either returning the image/SVG directly or saving it to disk

Build

npx tsc

Usage

Use with Claude desktop

{
  "mcpServers": {
    "mermaid": {
      "command": "npx",
      "args": ["-y", "@peng-shawn/mermaid-mcp-server"]
    }
  }
}

Use with Cursor and Cline

env CONTENT_IMAGE_SUPPORTED=false npx -y @peng-shawn/mermaid-mcp-server

You can find a list of mermaid diagrams under `./diagrams`, they are created using Cursor agent with prompt: "generate mermaid diagrams and save them in a separate diagrams folder explaining how renderMermaidPng work"

Run with inspector

Run the server with inspector for testing and debugging:

npx @modelcontextprotocol/inspector node dist/index.js

The server will start and listen on stdio for MCP protocol messages.

Learn more about inspector here.

Installing via Smithery

To install Mermaid Diagram Generator for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @peng-shawn/mermaid-mcp-server --client claude

Docker and Smithery Environments

When running in Docker containers (including via Smithery), you may need to handle Chrome dependencies:

1. The server now attempts to use Puppeteer's bundled browser by default

2. If you encounter browser-related errors, you have two options:

Option 1: During Docker image build:

Option 2: Use Puppeteer's bundled Chrome:

For Smithery users, the latest version should work without additional configuration.

API

The server exposes a single tool:

• `generate`: Converts Mermaid diagram code to a PNG image or SVG file
  • Parameters:
    • `code`: The Mermaid diagram code to render
    • `theme`: (optional) Theme for the diagram. Options: "default", "forest", "dark", "neutral"
    • `backgroundColor`: (optional) Background color for the diagram, e.g. 'white', 'transparent', '#F0F0F0'
    • `outputFormat`: (optional) Output format for the diagram. Options: "png", "svg" (defaults to "png")
    • `name`: Name for the generated file (required when CONTENT_IMAGE_SUPPORTED=false)
    • `folder`: Absolute path to save the image/SVG to (required when CONTENT_IMAGE_SUPPORTED=false)

The behavior of the `generate` tool depends on the `CONTENT_IMAGE_SUPPORTED` environment variable:

• When `CONTENT_IMAGE_SUPPORTED=true` (default): The tool returns the image/SVG directly in the response
• When `CONTENT_IMAGE_SUPPORTED=false`: The tool saves the image/SVG to the specified folder and returns the file path

Environment Variables

• `CONTENT_IMAGE_SUPPORTED`: Controls whether images are returned directly in the response or saved to disk
  • `true` (default): Images are returned directly in the response
  • `false`: Images are saved to disk, requiring `name` and `folder` parameters

Examples

Basic Usage

// Generate a flowchart with default settings
{
  "code": "flowchart TD\n    A[Start] --> B{Is it?}\n    B -->|Yes| C[OK]\n    B -->|No| D[End]"
}

With Theme and Background Color

// Generate a sequence diagram with forest theme and light gray background
{
  "code": "sequenceDiagram\n    Alice->>John: Hello John, how are you?\n    John-->>Alice: Great!",
  "theme": "forest",
  "backgroundColor": "#F0F0F0"
}

Saving to Disk (when CONTENT_IMAGE_SUPPORTED=false)

// Generate a class diagram and save it to disk as PNG
{
  "code": "classDiagram\n    Class01  Still\n    Still --> [*]\n    Still --> Moving\n    Moving --> Still\n    Moving --> Crash\n    Crash --> [*]",
  "outputFormat": "svg",
  "name": "state_diagram",
  "folder": "/path/to/diagrams"
}

FAQ

Doesn't Claude desktop already support mermaid via canvas?

Yes, but it doesn't support the `theme` and `backgroundColor` options. Plus, having a dedicated server makes it easier to create mermaid diagrams with different MCP clients.

Why do I need to specify CONTENT_IMAGE_SUPPORTED=false when using with Cursor?

Cursor doesn't support inline images in responses yet.

Publishing

This project uses GitHub Actions to automate the publishing process to npm.

Method 1: Using the Release Script (Recommended)

1. Make sure all your changes are committed and pushed

2. Run the release script with either a specific version number or a semantic version increment:

# Using a specific version number
   npm run release 0.1.4

   # Using semantic version increments
   npm run release patch  # Increments the patch version (e.g., 0.1.3 → 0.1.4)
   npm run release minor  # Increments the minor version (e.g., 0.1.3 → 0.2.0)
   npm run release major  # Increments the major version (e.g., 0.1.3 → 1.0.0)

3. The script will:

Method 2: Manual Process

1. Update your code and commit the changes

2. Create and push a new tag with the version number:

git tag v0.1.4  # Use the appropriate version number
   git push origin v0.1.4

3. The GitHub workflow will automatically:

Note: You need to set up the `NPM_TOKEN` secret in your GitHub repository settings. To do this:

1. Generate an npm access token with publish permissions

2. Go to your GitHub repository → Settings → Secrets and variables → Actions

3. Create a new repository secret named `NPM_TOKEN` with your npm token as the value

Badges

License

MIT