An organization off switch
Turn MCP off for the organization and it blocks every assistant, whatever each user’s own settings say.
Connect to LLMs
Ask Claude which screens are offline, or ask ChatGPT to create a device group, and Wallboard’s MCP server turns the request into API calls. Each assistant signs in as a real user. Admins decide whether AI access is allowed at all, and what each person’s assistant may read, change or delete, and the platform checks those rules on every request the assistant makes.
How it works
Open AI agent access under Settings, Preferences and General. Set the defaults that new users will inherit.
Claude Code, Cursor and ChatGPT only need your Wallboard server URL ending in /mcp. Claude.ai also needs an OAuth client.
The client opens a Wallboard login. From then on the assistant works as that user, under that user’s role.
Try show me all devices that are offline, or change the brightness to 80% on one device.
Turn MCP off for the organization and it blocks every assistant, whatever each user’s own settings say.
Choose whether a user’s assistant can only read, can create and update, or can also delete.
Destructive actions stay blocked until an admin turns them on for a named user. Super-admins start with MCP off.
MCP settings sit on top of normal roles. An assistant can be limited further, but never gets more than its user.
The backend and gateway apply the rules to real requests. A blocked action returns a clear reason, not a silent failure.
Set how many AI tokens each customer can use, and configure API rate limits in Preferences.
Any client that follows the MCP authorization spec can connect over Streamable HTTP. Claude Code, Cursor and ChatGPT register themselves, so setup is the URL and a login. ChatGPT needs a Plus, Pro, Business, Enterprise or Edu plan with Developer mode on, and a workspace admin must allow custom connectors on business plans. For Claude.ai, an admin first creates an OAuth client in Wallboard and pastes its ID and secret into Claude. Permission changes can take up to five minutes to reach an assistant.
Bring the questions your security review will ask. In the demo we’ll set a user to read-only, show an assistant being refused a change, then switch AI access off for the whole organization, and our support team will stay with you while you roll it out to a first pilot group.