How to Control Stream Deck with AI using MCP

Since Stream Deck 7.4, AI tools can connect to Stream Deck through MCP and trigger your actions by voice, text, or live events on your stream. NVIDIA Project G-Assist adds voice control, and Aitum Nexus lets raids and subscriptions set off your actions automatically.

In Stream Deck 8.0 beta, connecting takes a few clicks. Claude, ChatGPT, and other AI apps connect from a new AI tab, with no terminal or config files. Once connected, they can also place and configure actions on your keys for you.

MCP stands for Model Context Protocol. It is an open standard that gives AI tools a shared way to connect to apps and services.

Think of it like USB. Before USB, every device needed its own type of cable and connector. USB created one standard that works with everything. MCP does the same thing for AI tools. Instead of each AI assistant needing a custom connection to each app, MCP gives them a common way to communicate.

Thousands of apps and services support MCP, and Stream Deck joined them in version 7.4. That makes Stream Deck an AI controller: AI provides the input, and Stream Deck provides the actions.

MCP is at its best when you want to combine actions or stay hands-free.

With NVIDIA Project G-Assist, your voice tells Stream Deck what to do. Say “get my stream ready” and Project G-Assist can combine your actions to open OBS, set your scene, turn on your lights, and start your music. Your hands never leave the keyboard and mouse.

With Aitum Nexus, live events on your stream do the triggering instead of you. A raid comes in and your actions combine to switch scenes, dim the lights, and play a welcome sound. A channel point redemption activates a voice effect. A subscription kicks off a shoutout in Discord.

With Stream Deck 8.0 beta, AI apps like Claude and ChatGPT can also set up Stream Deck for you. Ask it to sort a full page of keys into rows by type, and it plans the layout and moves every key.

Stream Deck 8.0 beta builds MCP into Stream Deck app, so setup takes a few clicks instead of a terminal, and AI apps can set up Stream Deck as well as trigger actions.

In Stream Deck 8.0 beta, AI apps connect from the new AI tab. Node.js and the terminal aren’t needed. Download the beta from elgato.com/downloads.

1. Open Stream Deck app.
2. Go to Preferences and select the AI tab.
3. To trigger actions by voice or text, turn on Enable MCP Deck.
4. Find your AI app in the list and click Set Up.
5. Click Confirm. Stream Deck adds itself to the app’s settings and saves a backup of the original file.
6. Restart the AI app.

Your app shows Configured once it’s connected.

Stream Deck can set up these apps automatically:

• Claude Desktop
• ChatGPT & Codex
• Claude Code
• Cursor
• Visual Studio Code
• Windsurf
• Gemini CLI
• LM Studio

Stream Deck app needs to stay open while you use an AI app with it.

To control these apps from your keys too, see our guides for Claude, ChatGPT and Codex, and Cursor.

Turning on Enable MCP Deck adds a virtual device called MCP Deck to Stream Deck app, with 32 keys. AI apps can only trigger the actions you place on it, so add the ones you want to run by voice or text.

Each action on MCP Deck can have a description that tells AI apps what it does and when to use it. Select the action, click the AI icon, and describe it the way you’d ask for it, such as “This mutes my mic.” Descriptions matter most when you have similar actions, like two Multimedia actions that control different apps.

Some AI tools connect to a local address instead of appearing in the list. For those, turn on HTTP, then point the tool to:

To use a different port, click Configure… and change Port.

Once an AI app is connected in Stream Deck 8.0 beta, you can ask it to set up Stream Deck. It reads your devices, profiles, and installed plugins, then places and configures the actions itself.

AI apps can set up actions from plugins that support it. Most Elgato plugins already do, and other plugins add support as their makers update them.

On Stream Deck 7.4 or later without the beta, AI tools connect through Elgato MCP Server, which runs on Node.js.

1. Open Stream Deck app.
2. Go to Preferences and select the General tab.
3. Enable MCP Deck.

This adds a virtual device called MCP Deck to Stream Deck app, with 32 keys.

Drag actions onto its keys from the action panel on the right, just like you would with any other profile. Any action you place here becomes available to connected AI tools. Actions on your other profiles stay private, so you control exactly what the AI can access.

Each action on MCP Deck has a description field that tells the AI what the action does and when to use it. This is how the AI decides which action to trigger when you make a request.

To add a description, select an action, click the AI icon, and enter the phrases you’d naturally say.

Elgato MCP Server runs on Node.js, so you need to install it before connecting any AI tool. If you already have Node.js on your computer, you can skip this step.

The quickest way is to download the installer from nodejs.org. Select the LTS version, then run the installer and follow the prompts.

Windows: Download the .msi installer. Run it, click through the setup wizard, and make sure Add to PATH stays checked.

macOS: Download the .pkg installer. Open it and follow the prompts.

To confirm it worked, open a terminal (Terminal on macOS, or right-click the Start button and select Terminal on Windows) and run:

If you see a version number, Node.js is ready.

Elgato MCP Server is the bridge between AI tools and Stream Deck. It discovers the actions on MCP Deck and makes them available to any connected AI tool.

Some tools, like Claude Desktop, run Elgato MCP Server automatically in the background, so you can skip ahead to the connection section for that tool. Others, like NVIDIA Project G-Assist, need you to install and start it yourself.

1. Open a terminal: Terminal on macOS, or right-click the Start button and select Terminal on Windows.
2. Windows only: Allow npm to run by entering this command, then type Y and press Enter to confirm:

This changes the setting for your Windows user account only. Without it, PowerShell blocks npm and npx from running.

1. Install Elgato MCP Server:

This installs the server globally so you can start it anytime with a single command.

NVIDIA Project G-Assist brings voice control to Stream Deck using MCP. To get started, check out our guide to setting up NVIDIA Project G-Assist with Stream Deck.

In Stream Deck 8.0 beta, you can skip these steps: click Set Up next to Claude Desktop in the AI tab instead. On Stream Deck 7.4, Claude Desktop stores its MCP connections in a configuration file, so you add Elgato MCP Server to that file yourself.

1. Open Claude Desktop.
2. Go to Settings, then Developer, then click Edit Config. This opens the folder where the configuration file is stored.
3. Find claude_desktop_config.json in that folder. Right-click it and open it with a text editor (Notepad on Windows, TextEdit on macOS).
4. Add an mcpServers entry inside the existing structure, not as a separate block underneath. Find the last } that closes an existing section, add a comma after it, then add the mcpServers section before the final closing }.
5. Save the file and fully quit Claude Desktop (not just close the window). On macOS, right-click the dock icon and select Quit.

Here is what a typical file looks like after the edit:

Everything must stay inside one set of outer curly braces, with a comma between each section. If the mcpServers block gets pasted as a separate object below the existing content, Claude Desktop will show an error on startup.

If the file is empty or only contains {}, you can replace everything with:

To verify everything is working, go to Settings, then Developer. Under Local MCP servers, you should see elgato listed with a green running label.

If it does not appear or does not show as running, check that Stream Deck app is open with MCP Deck enabled and that the configuration file was saved correctly. Then fully quit and reopen Claude Desktop.

NVIDIA Project G-Assist, Aitum Nexus, and Claude Desktop are a few ways to connect, but because MCP is an open standard, any compatible tool can plug in the same way.

Elgato MCP Server supports two transport modes. stdio is the default and works with tools like Claude Desktop that manage the server process directly. HTTP runs the server on a local port and works with tools like Project G-Assist that connect over the network. You can also run it temporarily without a global install using npx:

For full setup details, additional connection methods, and advanced configuration, visit the Elgato MCP Server page on npm.

Stream Deck 8.0 beta: If your AI app can’t reach Stream Deck or an action doesn’t trigger, check the following:

1. Stream Deck app is open.
2. The AI tab shows Configured next to your app. If it shows Installed, click Set Up again.
3. You restarted the AI app after setting it up.
4. Enable MCP Deck is on, and the action you want to trigger is on MCP Deck.
5. If another program already uses port 29001, click Configure… and change Port.

Stream Deck 7.4: If an action does not trigger, check the following:

1. Stream Deck app is running and Enable MCP Deck is checked in Preferences.
2. The action is on MCP Deck, not on another device or profile.
3. Your AI tool is connected to Elgato MCP Server.
4. If you started Elgato MCP Server yourself (for NVIDIA Project G-Assist, for example), the terminal window running it is still open. Closing that window stops the server. To check that it’s running, open http://localhost:9090/health in a browser. A status response means the server is running and connected to Stream Deck app.
5. On Windows, if PowerShell says running scripts is disabled, run Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy RemoteSigned, type Y to confirm, then install Elgato MCP Server again.

The more actions and plugins you have set up in Stream Deck, the more a connected AI tool can do with them. Explore plugins on Elgato Marketplace to add more.