Run your screens from your AI assistant
Ask the assistant your team already uses which screens are offline, put a playlist on every screen at a site, or turn a page it just made into an app on your screens. Screenly has no AI of its own: you connect yours, and a token you control decides what it can touch.
Which screens in the Frankfurt plant are offline?
Two of the 14 screens labelled Frankfurt are offline: Line 3 andon since 08:12 and Gate B lobby since 07:55.
Put the Safety Rotation playlist on every screen labelled Warehouse.
Done. Safety Rotation now plays on the 23 screens labelled Warehouse.
Your assistant, on your terms
An assistant connects to Screenly through the Screenly MCP server, part of the Screenly CLI, running on your own computer.
- 01
No AI built into Screenly
Screenly sends nothing from your workspace to an AI model. An assistant sees only what it asks for, and only once you connect one yourself.
- 02
The assistant you already approved
Claude Desktop installs Screenly as an extension. Cursor and other assistants that support MCP connect with a short configuration.
- 03
Your token, your limits
The assistant holds no access of its own. It gets exactly what its API token allows, in one workspace, and deleting the token stops it at once.
Ask about your fleet in plain words
"Which screens in the Frankfurt plant are offline, and since when?" is one question to an assistant, answered as a sentence you can paste into a ticket. In the dashboard the same answer is a filter, a sort and a scan down the list.
Screens are read-only to an assistant. It can tell you a screen is offline, what it plays and what it last reported, but it cannot pair, change, reboot or delete one, so the hardware stays with the people who manage it.
Change many screens at once, through labels
Labels group screens and content, and they are the assistant's lever on a fleet just as they are yours. "Put the Safety Rotation playlist on every screen labelled Warehouse" is one request, where the dashboard would take a click per screen.
Playlists work the same way: add a video to the start of one, reorder it, or build a new one from content already in your library. Keep a request to a label rather than everything, and a misunderstood instruction reaches only one group.
Publish what your assistant just made
Ask an assistant for a page, such as a status board for four production lines, and it can publish that page to Screenly as an app, ready for a playlist. It runs on your screens like any other app.
Built for a security review
Every action an assistant can take is marked read-only, change or destructive, so an assistant that supports confirmations asks before it deletes a playlist rather than after. About half of them only read.
The real control is the token: one from an account with just the access the assistant needs, revoked when you are done. With the Claude Desktop extension it is kept in your operating system's keychain. Why Screenly has no built-in AI sets out the reasoning.
What an assistant can do in Screenly
Through the Screenly MCP server, with the access of your API token. Pairing, screen settings and app configuration stay in the dashboard.
Questions about your screens
Status, sync and hardware details for every screen, filtered by site, label or anything you ask about.
Working with an assistantContent and playlists
Add a web page, image or video from a link, build playlists and change what they play.
Tool referenceApps from a page
Publish a page the assistant made as a Screenly app, ready for a playlist.
Publish HTML as an appConfirmation before deleting
Every action is marked read-only, change or destructive, so an assistant can ask first.
What the MCP server isAccess set by a token
One workspace, exactly the access of the account that made the token, revoked in seconds.
PermissionsClaude Desktop, Cursor and more
An extension for Claude Desktop, and a configuration for Cursor and other MCP clients.
Connect a client
AI assistants in Screenly, answered
Does Screenly have AI built in?
No. Screenly does not send your content, screens or workspace to an AI model. If your team uses an assistant, you connect it yourself, and it sees only what it asks for.
Which AI assistants work with Screenly?
Any assistant that supports the Model Context Protocol (MCP). Claude Desktop installs Screenly as an extension; Cursor and other clients use a short configuration.
What can an assistant change?
Content, playlists, labels and apps, within the access of its token. It can read screens but not pair, change, reboot or delete them, and it works in one workspace at a time.
How do I keep an assistant from doing something I did not mean?
Give it a token from an account with only the access it needs, keep requests to a label, and turn on confirmations if your assistant supports them. Delete the token and the assistant stops.
Do I need to be a developer to use it?
No. With Claude Desktop you install the Screenly extension and add an API token from your workspace, about ten minutes the first time.