editor

This MCP server automates the PlayCanvas Editor workflow by exposing a set of Editor-related tools to a large language model. It organizes capabilities into entities, assets, scenes, stores, and viewport actions so the LLM can inspect, create, modify, and query assets and scene graphs, instantiate assets, apply scripts, and capture viewport images among other tasks. The server is designed to be driven by Claude (recommended) or Cursor, enabling natural language prompts to translate into structured MCP tool calls.

To use it, connect the PlayCanvas Editor to the MCP server via the supplied extension/config, then issue commands through Claude Desktop or Cursor. The server exposes tools such as listing, creating, deleting, and modifying entities and components; managing assets (listing, creating, deleting, and applying scripts or materials); querying and updating scene settings; performing store lookups and downloads; and taking viewport captures. Ensure your connected LLM stays within the allowed chat context and uses the Tools provided by the MCP Server to perform actions inside the PlayCanvas Editor.

Prerequisites:

• Node.js and npm installed on your machine
• Access to the PlayCanvas Editor and, optionally, the MCP extension setup described below

Installation steps:

1. Install dependencies for the MCP server project

npm install

2. Install or set up an MCP launcher (e.g., Cursor or Claude Desktop) per your environment

3. Ensure you have a working path to the server TS entry and adjust the MCP config accordingly. The README examples show using npx with tsx to run a TypeScript server entry, for example:

{
  "mcpServers": {
    "playcanvas": {
      "command": "npx",
      "args": [
        "tsx",
        "/path/to/editor-mcp-server/src/server.ts"
      ],
      "env": {
        "PORT": "52000"
      }
    }
  }
}

4. Start the MCP server using your configured command (adjust PATH as needed) and verify the port is reachable. If you are on Windows, you may use the equivalent cmd invocation shown in the README example. On macOS/Linux, the npx-based approach is typical.

Tips and caveats:

• The MCP server is designed to connect to either Claude Desktop or Cursor. Claude Desktop is generally more reliable according to the docs.
• In Cursor, enable auto-run mode in the Chat Features to allow the LLM to trigger MCP tools without frequent authorizations, but use this at your own risk.
• Ensure the MCP Config JSON port matches the port the server is listening on (default 52000 in the examples).
• Only one PlayCanvas Editor instance can connect to the MCP Server at a time via the extension, per the current guidance.
• If you run into connection issues, verify the Chrome extension is loaded (for the PlayCanvas Editor integration) and that the port in the MCP config matches the one the server exposes.