DevMotion MCP Server

Let AI assistants build DevMotion videos for you — create projects, add layers, animate, and then render through the CLI or hosted Studio.

There are two ways to connect, and local comes first:

• Local (stdio) — the default for devs & agents. The MCP server runs from the CLI, right inside your project. No account, no network, no tokens. This is what devmotion setup wires up, and it edits your HTML files directly.
• Remote (cloud HTTP) — optional. Point any MCP client at the hosted server when you’d rather work against cloud projects in the hosted Studio.

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI clients interact with external tools and services. By connecting your AI client to DevMotion’s MCP server, the AI can create and edit animation projects on your behalf.

Quick start (local)

Inside a DevMotion project (run devmotion init my-video first), run:

npx @devmotion/cli setup

This installs the Claude Code skills and registers a local, stdio MCP server in .claude/settings.local.json:

{
  "mcpServers": {
    "devmotion": {
      "command": "npx",
      "args": ["@devmotion/cli", "mcp"]
    }
  }
}

Open your agent and ask it to build a video. The AI calls upsert_layer, configure_project, and friends, and every change is written straight back to your HTML files. See AI Coding Assistants for Cursor, Windsurf, and manual config.

No account, no API key, no quota — the local server runs entirely on your machine.

Prefer the cloud? (remote HTTP)

If you’d rather work against cloud projects in the hosted Studio, connect to the remote endpoint instead:

MCP Endpoint: https://devmotion.app/mcp

Interactive clients (Claude, Cursor, Windsurf) handle OAuth automatically; programmatic clients use a bearer token from your profile page. See AI Coding Assistants for client configs and Enterprise & Automation for no-code platforms and agent frameworks.

Which one should I use?

How it works

1. Connect — point your client at the local stdio server (npx @devmotion/cli mcp) or the remote endpoint. The AI discovers available tools automatically.

2. Create — ask the AI to build an animation in plain language:

“Create a 1080×1920 TikTok intro with a bold title that slides in and a pulsing accent shape in the background.”

3. Render — locally with devmotion render, or via the hosted Studio when using the remote server. See the full tool reference.