Clip.A.Canvas MCP Server

A universal MCP (Model Context Protocol) server that exposes Clip.A.Canvas HTML-to-video rendering as an MCP tool. Works with any STDIO-based MCP client.

Features

• Universal MCP tool — render_video and render_video_to_file exposed via the MCP protocol
• Works everywhere — Claude Desktop, Codex, Gemini CLI, Qwen Coder, and any MCP-compatible AI coding tool
• Local rendering — no upload, no cloud; Chromium + FFmpeg run on your machine
• Base64 or file output — return video as base64 string or save directly to a path

Installation

One-command MCP install

After publishing to PyPI:

# Gemini CLI
gemini mcp add clipacanvas -- uvx clipacanvas-mcp

# Codex CLI
codex mcp add clipacanvas -- uvx clipacanvas-mcp

# Claude Code on Windows
claude mcp add -s user clipacanvas -- cmd /c uvx clipacanvas-mcp

From PyPI

pip install clipacanvas-mcp
# or
uv tool install clipacanvas-mcp  # gives 'clipacanvas-mcp' command

From source (editable)

cd code2video/mcp
pip install -e .

uvx handles the Python package environment. On first render, the MCP package will also install Playwright Chromium automatically if it is missing.

Configuration

Environment variables (all optional):

Variable	Default	Description
CLIPACANVAS_FFMPEG_EXE	auto-detect	Path to FFmpeg executable
CLIPACANVAS_BROWSERS_PATH	auto-detect	Path to Playwright browsers
CLIPACANVAS_MAX_DURATION	12	Max render duration in seconds
CLIPACANVAS_DEFAULT_WIDTH	540	Default video width
CLIPACANVAS_DEFAULT_HEIGHT	960	Default video height
CLIPACANVAS_DEFAULT_BITRATE	5M	Default FFmpeg bitrate
CLIPACANVAS_DEFAULT_FRAME_RATE	60	Default frame rate

MCP Tools

render_video

Render HTML/CSS/JS to MP4, returned as base64.

Arguments:

{
  "code": "<!DOCTYPE html><html>...</html>",
  "width": 540,
  "height": 960,
  "bitrate": "5M",
  "frame_rate": 60,
  "max_duration": 12
}

Returns: Text content with video metadata and base64-encoded MP4 data.

render_video_to_file

Same as above but saves the MP4 directly to output_path.

Arguments:

{
  "code": "<!DOCTYPE html><html>...</html>",
  "output_path": "/absolute/path/to/video.mp4",
  "width": 540,
  "height": 960
}

Usage in AI Coding Tools

Claude Code

For this local repo checkout:

cd code2video/mcp
pip install -e .
claude mcp add -s local clipacanvas -- cmd /c python -m clipacanvas_mcp.server

The cmd /c wrapper is useful on Windows because Claude Code's MCP health check can fail to resolve the same Python command that works in an interactive PowerShell session.

Claude Desktop

Add to your Claude Desktop config file:

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

{
  "mcpServers": {
    "clipacanvas": {
      "command": "python",
      "args": ["-m", "clipacanvas_mcp.server"]
    }
  }
}

Then restart Claude Desktop. The render_video and render_video_to_file tools will appear in the tool list.

Codex CLI

For this local repo checkout:

cd code2video/mcp
pip install -e .
codex mcp add clipacanvas -- python -m clipacanvas_mcp.server

Gemini CLI (Google)

For this local repo checkout, install the server in editable mode first:

cd code2video/mcp
pip install -e .
gemini mcp add clipacanvas -- python -m clipacanvas_mcp.server

Use the uvx command only after clipacanvas-mcp is published to a package index and uvx is installed:

gemini mcp add clipacanvas -- uvx --from clipacanvas-mcp clipacanvas-mcp

Qwen Coder (Alibaba)

Add to Qwen Coder's MCP configuration:

{
  "mcpServers": {
    "clipacanvas": {
      "command": "uvx",
      "args": ["--from", "clipacanvas-mcp", "clipacanvas-mcp"]
    }
  }
}

Any STDIO MCP Client

# Direct run
uvx --from clipacanvas-mcp clipacanvas-mcp

# Or after pip install
clipacanvas-mcp

Testing the Server Manually

# Start the server
python -m clipacanvas_mcp.server

# Send initialize
echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}' | clipacanvas-mcp

# List tools
echo '{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}' | clipacanvas-mcp

Requirements

• Python 3.10+
• FFmpeg (installed system-wide, via imageio-ffmpeg, or set CLIPACANVAS_FFMPEG_EXE)
• Playwright Chromium: installed automatically on first render if missing; run python -m playwright install chromium manually only if you want to preinstall it

License

MIT