← All use cases

Edit an LED and DMX timeline with Show Designer

LightingIntermediateShow DesignerArt-NetsACNDMXMCP

What it does

Show Designer Pro is a lighting program for LED strips and DMX fixtures. The engine is headless Python. A React page in the browser is the timeline, the patch, the live view and a Three.js preview. The README says audio plays on the PC as the master clock, and that the same backend exposes JSON-RPC so an assistant can drive the show.

Documented outputs are Art-Net (per-universe unicast), sACN E1.31 (unicast or multicast) and an ENTTEC Open DMX USB device via pyserial. Fixture profiles are GDTF or JSON. Routing from a universe to a target lives in output_targets.json, separate from the rig. The README's quick start is Windows 10 or 11 with Python 3.11 or newer. It says Linux and macOS are not guaranteed. Node 18 or newer is only needed if you rebuild the frontend. The built UI is already in web/dist.

The repo's .mcp.json starts python -m src.mcp.mcp_show_server over stdio. That module connects to ws://127.0.0.1:9876 and forwards each tool as JSON-RPC. server/main.py describes that port as the MCP-compatible bridge. On 21 October 2026 the module marked 62 functions with the MCP tool decorator. They cover transport (play, pause, stop, seek, blackout, get_state), clips, channel clips, fixtures and the rig, cue slots, groups, saving and loading a show, and audio-analysis helpers such as analyzer_find_drops. generate_section takes a start, a duration and a style. The MCP guide's style names are energetic, subtle, synchronized, random and audio_reactive.

The project README describes the same bridge as 150 or more JSON-RPC commands. The MCP guide says 50 or more tools. The count checked in src/mcp/mcp_show_server.py is the 62 decorators above. The license in the repo is Prosperity Public License 3.0: the README says personal, educational and open-source use is free, and commercial use needs a license, with a 30-day trial stated there.

How to set it up

  1. On Windows, clone the repo, create a virtualenv with Python 3.11 or newer, activate it, and pip install -r requirements.txt. Start the app with python -m server.main. The README says the UI is then at http://localhost:8000. Luces.bat is the Windows launcher that restarts cleanly and opens the browser.
  2. Leave that process running. The stdio MCP server only forwards calls. If nothing is listening on port 9876, a tool returns a bridge error instead of editing the show. The MCP guide's connection-refused note says the app crashed or port 9876 is already taken.
  3. Claude Code, opened in the repo, reads .mcp.json. The server name there is show-control, the command is python, and the args are -m and src.mcp.mcp_show_server. The MCP guide says to start Claude Code after Show Designer is already up, and that the tools show up as mcp__show-control__*.
  4. Ask for a read first, such as the current state or the fixture list, before add_clip or generate_section. blackout and play hit the running show. save_show writes it.
  5. Output targets are not the same file as the rig. Edit output_targets.json for which universe goes to WLED, an Art-Net node, sACN or the USB widget. The hardware guide linked from the README is docs/hardware.md.

Source project: guillegar/show_designer

FAQ

Does the assistant talk to the fixtures by itself?

No. The stdio process opens a WebSocket to 127.0.0.1 port 9876 and the Show Designer process does the Art-Net, sACN or USB output. Stop the app and the tools have nothing to forward to.

Which tool count should I plan context around?

Use the MCP server you actually launch. The file .mcp.json points at src.mcp.mcp_show_server, which currently decorates 62 tools. The README's "150+" figure is about JSON-RPC commands on that bridge, and the MCP guide's figure is "50+".

Can I use this on a paid gig?

The README says Prosperity Public License 3.0 allows personal, educational and open-source use, and that commercial use needs a license. It also says that license includes a 30-day trial. Read LICENSE in the repo before a paid show.

Original source: https://github.com/guillegar/show_designer

AI on Stage, a weekly newsletter

New agent use cases for live events, once a week.