Scribe MCP server

Let your AI agent make hand-drawn videos.

Todrawn Scribe has an MCP server. Claude Code, Claude Desktop, Cursor, VS Code, Codex or any other MCP client can turn code, a sketch or a photo into a Scribe project and hand you a link. You watch the pen draw it and export the video in your browser.

Install in Claude Code, Claude Desktop and more

Claude Code

claude mcp add todrawn-scribe -- npx -y @todrawn/scribe-mcp

OpenAI Codex CLI

codex mcp add todrawn-scribe -- npx -y @todrawn/scribe-mcp

VS Code (GitHub Copilot)

code --add-mcp '{"name":"todrawn-scribe","command":"npx","args":["-y","@todrawn/scribe-mcp"]}'

Claude Desktop, Cursor and Gemini CLI

{
  "mcpServers": {
    "todrawn-scribe": {
      "command": "npx",
      "args": [
        "-y",
        "@todrawn/scribe-mcp"
      ]
    }
  }
}

Add this block to the client's MCP config. Claude Desktop: Settings > Developer > Edit Config. Cursor: ~/.cursor/mcp.json. Gemini CLI: ~/.gemini/settings.json. Then restart the client.

Any other MCP client starts the same command over stdio: npx -y @todrawn/scribe-mcp

Needs Node.js 20 or newer. No API key, no account and no configuration: the server runs on your machine.

What your agent can do

Code and text to video

A function, a diff or a few paragraphs. Scribe writes it out by hand with syntax colours, at reading pace.

Try asking: Make a vertical Scribe video of the function I just wrote.

Line drawings

The agent describes the picture as SVG paths, which models are good at, and the pen draws them stroke by stroke in the order the agent chose.

Try asking: Draw a lighthouse at night with Scribe.

Photos, painted

Point it at a PNG or JPEG. The server paints it on your machine into up to about 2,800 coloured strokes, an underpaint first and detail last: the method behind our own Scribe ads.

Try asking: Paint ~/Pictures/dinner.jpg with Scribe.

See what the link opens

Both links below came straight from the server. Each opens the studio with the project loaded; watching needs no account.

Open the code demo Open the drawing demo

Nothing is uploaded

The server only builds links. The whole project, even a painted photo's strokes, is packed into the part of the link after #, which browsers never send to a server. Your code and photos stay on your machine until you choose to export or share.

Limits and plans

Watching and editing the project is free and needs no account. Exporting the MP4 needs a free Todrawn account: the Free plan exports 5 videos a day, each ending with a short Todrawn card, and Premium exports without limits. Clips are capped at 180 seconds; longer projects are drawn faster to fit.

Questions

Which AI tools does it work with?

Any client that speaks the Model Context Protocol over stdio. The most popular ones have a ready snippet above: Claude Code, Claude Desktop, Cursor, VS Code with GitHub Copilot, OpenAI Codex and Gemini CLI. Others, such as Cline or Zed, take the same npx command.

Do I need an API key or a Todrawn account?

Not to install it or to watch what it makes. An account is only needed when you export the video, the same as in the studio.

Can the agent download the MP4 itself?

Not yet. The agent hands you a link, and the video is rendered and exported in your browser. That keeps your files local and your export limits exactly as they are in the studio.

What kind of photos paint well?

Clear subjects with strong shapes of colour: food, portraits, landscapes, products. Fine texture such as text in a photo or tiny leaves turns into dabs, because the pen is only a few pixels wide.

Can I change the result?

Yes. The link opens an ordinary Scribe project: change the pen, speed, background, aspect or the strokes themselves before you export.

Prefer to do it by hand?

Everything the agent does, you can do yourself in the studio, in your browser.

Open Todrawn Scribe