Backdraw
Generate images through your own signed-in ChatGPT in a background Chrome tab on Windows
- Pricing
- Open source
- Platforms
- Windows, Chrome extension, CLI
- Runs
- Locally
- Version
- 2.4.2
- Listing updated
- today
Similar MCP servers
About Backdraw
Backdraw is an open-source MCP server for design and media. It runs locally on your machine and is started by your AI client.
Backdraw pairs a local MCP server with a Chrome extension. When you ask Claude for an image made with ChatGPT, the extension opens chatgpt.com in a muted background tab of your everyday Chrome, inside a ChatGPT Project you pick, sends the prompt, waits for the picture and saves it to Pictures under Backdraw. Claude gets the file path, a preview and the chat link.
It can edit an image attached as a reference, ask for aspect ratios such as 16:9 or a 4:1 banner, request transparent backgrounds, and build animated GIFs by looping a PNG or aligning a ChatGPT sprite sheet. The sprite step checks frame count, empty frames and character identity. A right-click collects reference images from any website, and a gallery side panel browses past work.
There is no API key, so each image counts against your ChatGPT plan limits. Pieces talk over Chrome native messaging and a Windows named pipe, with no open port or Backdraw server. The README warns that automating the ChatGPT page may conflict with OpenAI's terms. It needs Windows 10 or 11, Chrome 116 or newer and Node.js 20.
Install
Run in a terminal
claude mcp add backdraw -- npx -y backdrawAdd to claude_desktop_config.json
{
"mcpServers": {
"backdraw": {
"command": "npx",
"args": [
"-y",
"backdraw"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"backdraw": {
"command": "npx",
"args": [
"-y",
"backdraw"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"backdraw": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"backdraw"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"backdraw": {
"command": "npx",
"args": [
"-y",
"backdraw"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"backdraw": {
"command": "npx",
"args": [
"-y",
"backdraw"
]
}
}
}Using several servers? Build one config file with the MCP config generator.
Tools this server provides (3)
Taken from the project documentation; the running server may differ.
- animate_image
- Loops a single PNG into an animated GIF locally.
- generate_sprite
- Generates a sprite sheet through ChatGPT.
- animate_sheet
- Turns a sprite sheet into an aligned transparent GIF.
What it does
- Background ChatGPT tab
- Image jobs run in a muted tab that never takes focus.
- Reference collections
- Right-click images on any site and reuse them as style references.
- Sprite and GIF animation
- Cut and align sprite sheets into transparent GIFs with quality checks.
- Gallery side panel
- Browse everything generated in Chrome's side panel.
- No open ports
- Uses native messaging and a user-only named pipe, not a DevTools port.
- One-command setup
- npx backdraw@latest setup configures clients and the native host.
Good for
- Make a hero image from Claude Code without an image API key
- Create a transparent-background sprite for a game
- Turn a sprite sheet into an animated GIF
- Generate images in the style of references saved from the web
Pricing
Open source
Backdraw FAQ
Does it need an API key?
No. It uses the ChatGPT session already signed in to your Chrome.
Which systems work?
Windows 10 or 11 only, with Chrome 116 or newer.
Where are images saved?
In Pictures under Backdraw, named by time and prompt.
Are there terms risks?
The README says automating ChatGPT may conflict with OpenAI's terms and use is at your own risk.
Is Backdraw free?
Yes. Backdraw is open source, so you can use and self-host it for free.
How do I add Backdraw to Claude Code?
Run this command in your terminal: claude mcp add backdraw -- npx -y backdraw
What are the best alternatives to Backdraw?
The closest alternatives to Backdraw are HyperFrames by HeyGen, next-ai-draw-io, mcp-for-blender, ffmpeg-skill and cursor-talk-to-figma-mcp.
Where this comes from
- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.



