Editor Mcp Server
MCP Server for AI automation of the PlayCanvas Editor
| npm | Developer Site | Editor Manual | API Reference | Blog | Forum |
An MCP server that lets AI agents automate the PlayCanvas Editor and check that scenes actually run. The MCP client is built into the Editor, so no browser extension is needed.
Install
Add the server to your MCP client. For Claude Code:
claude mcp add playcanvas -- npx -y @playcanvas/editor-mcp-server
Requires Node.js 22.18+. Add the server once per agent: two servers take turns owning the Editor port.
Claude Desktop, Cursor, and a shared project .mcp.json
{
"mcpServers": {
"playcanvas": {
"command": "npx",
"args": ["-y", "@playcanvas/editor-mcp-server"]
}
}
}
Add this to claude_desktop_config.json (Settings > Developer > Edit Config), to Cursor's
global MCP servers (Cursor Settings > MCP), or to a .mcp.json committed to the project root
for Claude Code. On Windows, use "command": "cmd" with "args": ["/c", "npx", "-y", "@playcanvas/editor-mcp-server"].
Codex CLI and app
codex mcp add playcanvas -- npx -y @playcanvas/editor-mcp-server
On Windows, insert cmd /c before npx. If the first run times out while npx downloads, raise
startup_timeout_sec under [mcp_servers.playcanvas] in ~/.codex/config.toml.
Then open your project in the Editor, click the MCP button below Publish, and click CONNECT. To
start with no project open, use the MCP button next to NEW PROJECT on playcanvas.com/editor. Allow
Chrome's Apps on device permission when prompted; if the Editor stays on Connecting, it is
blocked. The default port is 52000; to change it, pass --port <number> to the server and set the
same port in the Editor.
Usage
Ask your agent for the outcome you want in the open project. For example:
Add a spinning cube with a rotate script to the current scene, launch it, and capture a screenshot
of the running app to confirm it turns.
Coding agents: start from the Editor guides index and the Editor API index, which link a Markdown version of every page.
Tools
From the project's README.