Stream Deck is already the command center for your setup. Now you can trigger those actions with your voice using NVIDIA Project G-Assist, a local AI assistant that runs directly on your GeForce RTX graphics card.
Say "turn on my lights and launch OBS" and it happens. Say "clip that" and your replay buffer saves. Say "set me to BRB" and your entire be-right-back routine fires off in one go.
This is possible through MCP, or Model Context Protocol. A standard that lets AI agents see what actions are available in an app and trigger them based on your request.
What you need
- NVIDIA GeForce RTX GPU with at least 6GB VRAM
- A Stream Deck device or Stream Deck Mobile Pro
- Stream Deck app 7.4.1 or later
Video guide
1. Download Stream Deck app
Download Stream Deck app from elgato.com/downloads.
Note: Project G-Assist requires Stream Deck 7.4.1 or newer.
2. Install NVIDIA Project G-Assist
Open the NVIDIA App, go to the Home tab, find Project G-Assist under Discover, and click Get. Once installed, bring up Project G-Assist anytime by pressing ALT+G, or hold ALT+V to use voice input. The Stream Deck plugin comes pre-installed with Project G-Assist v0.2.1 and later.
3. Install Node.js
Open a terminal by right-clicking the Start button and selecting Terminal (Admin). Paste the following command below and press Enter:
winget install -e --id OpenJS.NodeJS.LTS
Wait for the installation to finish before moving to the next step.
4. Install Elgato MCP Server
In the same terminal window, paste this command below and press Enter:
npm install -g @elgato/mcp-server
5. Start Elgato MCP Server
Once the installation is complete, start the server by running:
elgato-mcp-server --http
Keep this terminal window open or minimized. The server needs to stay running while you use Project G-Assist with Stream Deck.
You will need to run this command again each time you start or restart your computer. To make this easier, paste the command into a batch file and save it anywhere on your computer, or download the ready-made file below. Then drag the batch file onto a Stream Deck key to create a shortcut. Tap it once and the server launches in the background automatically.
Download
6. Enable MCP Actions in Stream Deck
Open Stream Deck app and enable MCP Actions:
- Go to Preferences
- Select the General tab
- Check Enable MCP Deck
This creates a dedicated canvas called MCP Action in your device dropdown. This is where you set up everything Project G-Assist can control.
7. Build your MCP Deck canvas
Switch to the MCP Deck canvas and drag over any actions you want Project G-Assist to trigger. Switch scenes or save a clip in OBS Studio. Change your Twitch title or category while you're live. Toggle your lights. Fire off a Multi Action that gets your entire stream ready to go.
Any action available on a Stream Deck key, including plugin actions, can go on this canvas. Every action you add is another thing Project G-Assist knows how to do for you.
Note: Project G-Assist can only see and interact with actions on the MCP Deck canvas. It has no access to actions on your other Stream Deck profiles.
8. Add descriptions to your actions
Stream Deck is already the command center for your setup. Now you can trigger those actions with your voice using NVIDIA Project G-Assist, a local AI assistant that runs directly on your GeForce RTX graphics card.
Say "turn on my lights and launch OBS" and it happens. Say "clip that" and your replay buffer saves. Say "set me to BRB" and your entire be-right-back routine fires off in one go.
This is possible through MCP, or Model Context Protocol — a standard that lets AI agents see what actions are available in an app and trigger them based on your request. Stream Deck 7.4 adds MCP support, and NVIDIA Project G-Assist is built to use it.
Try it out
Once everything is set up, hold ALT+V and tell Project G-Assist what you want your Stream Deck to do. Every action you add to your canvas is something new it can handle.
For more information on Stream Deck MCP and other tools you can connect, check out how to Control Stream Deck with AI.