@pictify/mcp-server

@pictify/mcp-server

An MCP (Model Context Protocol) server for Pictify — generate images, GIFs, and PDFs from AI agents like Claude, Cursor, and Windsurf.

One-line install. No code required. Ask your AI assistant to create OG images, social media cards, screenshots, animated GIFs, PDF invoices, certificates, and more — all from natural language.

What can it do?

• Generate images from HTML/CSS, URLs, or reusable templates (OG images, Twitter cards, banners, product screenshots)
• Create animated GIFs from CSS animations or by recording live web pages
• Render PDFs from templates — invoices, certificates, reports, shipping labels
• Batch generate up to 100 personalized images in one request (team badges, event passes, product catalogs)
• A/B test images with built-in experiments, traffic splitting, and auto-optimization
• Template system with 50+ expression functions for dynamic content (conditionals, string manipulation, date formatting)

Works with Claude Desktop, Claude Code, Cursor, Windsurf, and any MCP-compatible client.

Quick Start

Option 1: Streamable HTTP with OAuth (Recommended)

No API key needed. The server handles authentication via your browser.

For MCP clients that support Streamable HTTP transport, use this URL:

http://localhost:3000/mcp

Start the server:

npx pictify-mcp-http

On the first request, your browser will open to sign in to Pictify. After that, everything works automatically.

Option 2: Stdio with API Key

For MCP clients that use stdio transport (most desktop apps), you'll need an API key:

• Sign up or log in at pictify.io
• Go to API Tokens
• Create a new token and copy it

Claude Desktop

Add to your config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "pictify": {
      "command": "npx",
      "args": ["-y", "@pictify/mcp-server"],
      "env": {
        "PICTIFY_API_KEY": "your_api_key"
      }
    }
  }
}

Restart Claude Desktop after saving.

Claude Code

claude mcp add pictify -e PICTIFY_API_KEY=your_api_key -- npx -y @pictify/mcp-server

Cursor

Add to Cursor's MCP settings (Settings > MCP Servers):

{
  "mcpServers": {
    "pictify": {
      "command": "npx",
      "args": ["-y", "@pictify/mcp-server"],
      "env": {
        "PICTIFY_API_KEY": "your_api_key"
      }
    }
  }
}

Windsurf

Add to Windsurf's MCP settings:

{
  "mcpServers": {
    "pictify": {
      "command": "npx",
      "args": ["-y", "@pictify/mcp-server"],
      "env": {
        "PICTIFY_API_KEY": "your_api_key"
      }
    }
  }
}

Available Tools

Image Generation

Tool	Description
pictify_create_image	Generate an image from HTML/CSS, a URL screenshot, or a template
pictify_create_canvas_image	Generate an image from FabricJS canvas JSON data
pictify_list_images	List previously generated images
pictify_get_image	Get details of a specific image by ID

GIF Creation

Tool	Description
pictify_create_gif	Create animated GIF from HTML with CSS animations
pictify_capture_gif	Record a GIF from a live web page over time
pictify_list_gifs	List previously generated GIFs
pictify_get_gif	Get details of a specific GIF by ID

PDF Generation

Tool	Description
pictify_render_pdf	Generate single-page PDF from a template
pictify_render_multi_page_pdf	Generate multi-page PDF from a template
pictify_list_pdf_presets	List available PDF page size presets

Template Management

Tool	Description
pictify_list_templates	List saved templates with filtering and pagination
pictify_get_template	Get template details
pictify_get_template_variables	Get template variable definitions and types
pictify_render_template	Render a template with variables (supports layout variants)
pictify_create_template	Create a new template (HTML or FabricJS)
pictify_update_template	Update an existing template
pictify_delete_template	Delete a template

Batch Operations

Tool	Description
pictify_batch_render	Start batch render job (up to 100 items, async)
pictify_get_batch_results	Check batch job status and get result URLs
pictify_cancel_batch	Cancel a running batch job

A/B Testing & Experiments

Tool	Description
pictify_list_experiments	List experiments (A/B tests, smart links, scheduled)
pictify_create_experiment	Create an experiment with variants and traffic weights
pictify_get_experiment	Get experiment details and per-variant performance
pictify_get_experiment_quota	Check experiment usage and plan limits
pictify_update_experiment	Update experiment config (field access depends on status)
pictify_delete_experiment	Delete an experiment
pictify_start_experiment	Start routing traffic to variants
pictify_pause_experiment	Pause traffic routing (data preserved)
pictify_complete_experiment	Declare a winner and route all traffic to it
pictify_track_experiment_events	Track impressions, clicks, and conversions

Configuration

Stdio Mode

Variable	Description	Default
PICTIFY_API_KEY	Your Pictify API key (required)	—
PICTIFY_BASE_URL	Custom API base URL	https://api.pictify.io
PICTIFY_DEBUG	Enable verbose logging to stderr	false

HTTP Mode (OAuth)

Variable	Description	Default
PICTIFY_BASE_URL	Pictify API base URL	https://api.pictify.io
PICTIFY_AUTH_SERVER_URL	OAuth authorization server URL	https://api.pictify.io
MCP_PORT	HTTP server port	3000

Examples

Create a social media card:

"Create a Twitter card image for my blog post titled 'Getting Started with MCP' with a blue gradient background, 1200x630."

Screenshot a website:

"Take a screenshot of stripe.com at 1440x900."

Render a template:

"List my templates and render the blog-header template with title 'Hello World'."

Batch generate images:

"Use my team-badge template to generate images for these 10 team members: ..."

A/B test an image:

"Create an A/B test experiment with two variants of my hero banner and start routing traffic."

Development

git clone https://github.com/pictify-io/pictify-mcp.git
cd pictify-mcp
npm install
npm run build

Test with MCP Inspector:

PICTIFY_API_KEY=your_key npm run inspector

Run HTTP mode locally:

npm run start:http

License

MIT