IlluGen MCP server

IlluGen MCP

Let your AI agent build, edit and render JangaFX IlluGen node graphs.

Describe a texture, a VFX flipbook or a procedural mask in plain words: your agent creates the .illugen project, wires the nodes, checks it and renders it with your own IlluGen. No clicks, no window.

Quick start · What it can do · Tools · How it works · FAQ

Real tool calls and timings, real IlluGen renders. Reproduce them with examples/demo_scenes.py.

Quick start

1. Requirements

include or replace IlluGen).

• Windows with IlluGen 1.2.x installed and activated (a valid JangaFX license: this project does not

• Python 3.10+. uv is recommended (no manual install needed).

2. Add it to your agent

Claude Desktop: one click

Download illugen-mcp-.mcpb from the latest release and open it (or drag it into Claude Desktop › Settings › Extensions). Needs uv on your PATH. If IlluGen is installed elsewhere, set its path in the extension settings.

Claude Code

claude mcp add illugen -s user -- uvx --from git+https://github.com/dokaNc/illugen-mcp illugen-mcp

Claude Desktop, Cursor, Windsurf, VS Code, Gemini CLI… (any MCP client): add this to the client's MCP configuration (claudedesktopconfig.json, .cursor/mcp.json, …):

{
  "mcpServers": {
    "illugen": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/dokaNc/illugen-mcp", "illugen-mcp"]
    }
  }
}

Without uv (plain pip)

pip install git+https://github.com/dokaNc/illugen-mcp
claude mcp add illugen -s user -- illugen-mcp

or "command": "illugen-mcp" in the JSON configuration.

IlluGen installed somewhere else?

Set ILLUGEN_EXE to the full path of IlluGen.exe:

claude mcp add illugen -s user -e ILLUGEN_EXE="D:\Apps\IlluGen\IlluGen.exe" -- uvx --from git+https://github.com/dokaNc/illugen-mcp illugen-mcp

"env": { "ILLUGEN_EXE": "D:\\Apps\\IlluGen\\IlluGen.exe" }

3. Ask

Restart your agent, then try:

Check that IlluGen is reachable with illugen_status, then build me a tileable stone mask
and render it to C:\Users\me\Desktop\stone.

What it can do

Shortened. The full README is on GitHub.