You automate digital signage updates with an AI assistant by connecting Claude, ChatGPT, Microsoft Copilot or Gemini to your signage CMS through a Model Context Protocol (MCP) server or a Zapier bridge. The assistant reads what each screen is playing, drafts the change you asked for, shows you a preview and waits for your confirmation before anything goes live. As of September 1, 2026, CastHub, Revel Digital, Screenly and Yodeck publish MCP servers, Fugo and NoviSign offer beta or bridged paths, and Skoop ships an in-product chat assistant.

This page covers which vendors support assistant control today, how the connection works, what still needs a person, and what it costs. It sits under the broader AI digital signage guide and the CastHub MCP overview.

What is an AI assistant for managing signage networks?

An AI assistant for managing signage networks is a general chat assistant such as Claude, ChatGPT or Microsoft Copilot that has been given tools to read and change a signage CMS. Rather than a chatbot bolted onto a dashboard, the assistant calls the vendor API through typed tool definitions, so it can list screens, check which devices are offline, assign a presentation, change a schedule or start an alert from one plain sentence.

Two things share the label. In-product AI, like Skoop Bot, the Ask Fugo sidebar or the Revel Digital Account Assistant, lives inside the vendor app. Assistant control connects the assistant you already use to the CMS from outside, so the Claude window that drafts your announcement can also publish it. CastHub takes the second approach and exposes 58 tools across 10 domains to any MCP-compatible assistant.

Which digital signage platforms work with AI assistants?

As of September 1, 2026, four platforms publish a vendor-run MCP server that Claude, ChatGPT or another MCP client connects to: CastHub, Revel Digital, Screenly and Yodeck. Fugo describes its MCP server as coming soon, NoviSign documents an MCP integration with a setup guide, and Skoop, ScreenCloud and OptiSigns offer in-product AI or Slack feeds without a published assistant control path. The table records what each vendor page said on the fetch date.

Vendor	Assistant path	MCP server	Published tool count	Governance	AI pricing
CastHub	Claude, ChatGPT, Microsoft Copilot, Gemini and Cursor via the hosted MCP server at webapi.cast-hub.com/mcp	Yes, production, hosted, OAuth 2.0 with PKCE	58 tools across 10 domains	Preview and confirmation before every write; ReadOnly, Destructive and Idempotent annotations. Audit logging and rollback on the roadmap, not shipped	No AI add-on fee; included on Basic ($12/mo, up to 2 screens) and Pro ($35/mo, up to 10 screens)
Revel Digital	Hosted MCP server for Claude, ChatGPT, Cursor, VS Code and Windsurf; in-product Account Assistant and agent tasks	Yes, hosted at mcp.reveldigital.io, browser OAuth sign-in	91 tools (page reads “Browse all 91 tools”)	ISO 27001; OWASP LLM Top 10 risk review; agent actions opt-in, described as auditable	All AI features on every plan; token usage allocated per managed player
Screenly	MCP server for Claude Desktop and Cursor, API token in the client config	Yes, API token authentication	33 tools across 8 categories, counted from the page	Inherits the API token permissions; no preview or confirmation step described	Not published on the MCP page
Fugo	Ask Fugo sidebar (beta), Actions with confirmation, AI triggers, MCP server for Claude and ChatGPT	Coming soon; OAuth 2.1 named	Not published	Confirmation step on Actions; stays within existing account permissions	Progressive rollout; Ask Fugo beta on select plans
NoviSign	In-Studio AI image creator plus an MCP integration with a setup guide	Described as an MCP integration; scope not on the page	Not published	Not published	Not published; 30-day free trial
Yodeck	AI agent through MCP for Claude, ChatGPT, Cursor, VS Code, Windsurf and Claude Code	Yes, OAuth 2.1 sign-in	Not published	Within existing Yodeck permissions; deletes disabled; tokens expire, never logged	Premium and Enterprise plans, no extra charge
ScreenCloud	QuickPost in-product AI content and Canva integration; Slack app mirrors channel posts	Not published	Not published	Not published	Not published
OptiSigns	Slack app shows channel messages on screens as a Slack Wall; no assistant control path	Not published	Not published	Wall filters (all, pinned, @OptiSigns) and moderation	Not published
Skoop	Skoop Bot in-product chat assistant (rename screens, publish, manage users) and AI content generation	Not published	Not published	Role-based access and activity logs at platform level	Not published; early adopter form

Tool counts are the vendor’s published figures, not a like-for-like measure of depth; a dashboard summary and a device restart each count as one. The digital signage MCP vendor guide goes deeper on architecture and authentication for each server.

How do you control digital signage with Claude?

You control digital signage with Claude by adding your signage vendor’s MCP server as a connector in Claude.ai or Claude Desktop, signing in through the vendor’s OAuth page, and asking for the change in plain language. With CastHub you paste the connector URL, approve the OAuth prompt, and Claude can list your devices, preview a schedule change and publish it after you confirm. Setup takes a few minutes and needs no code.

The connect Claude to CastHub guide walks through the connector screens. A good opening request is read-only, such as which devices are offline and what the lobby group plays this week. Read tools run without a confirmation prompt, so you see how Claude interprets your account before you let it write. Revel Digital and Yodeck document the same flow; Screenly’s server runs locally through its CLI with an API token.

Can ChatGPT control digital signage?

Yes. ChatGPT controls digital signage through custom connectors on paid ChatGPT plans, which attach to the same MCP servers Claude uses. CastHub, Revel Digital and Yodeck each name ChatGPT as a supported client on their pages as of September 1, 2026, and Fugo names it for its planned MCP server. You add the connector, sign in with OAuth, and ChatGPT calls the same tools with the same preview and confirmation behavior.

The connect ChatGPT to CastHub guide covers the connector settings, and the Microsoft Copilot guide covers Copilot Studio. The practical difference between assistants is small, because the tools, the previews and the confirmation prompts come from the CMS side, not from the assistant.

How does natural language control for digital signage work?

Natural language control works in five steps. The assistant reads your sentence, picks the tools that match it, calls read tools to learn the current state, drafts the write calls and shows you a preview, then executes after you confirm. A request such as “move the lobby group to the holiday-hours presentation from Friday through Monday” becomes a device lookup, a presentation lookup, a schedule assignment and a verification read, with the preview in between.

You never type tool names, but they are ordinary API operations with typed inputs. In CastHub the sentence above maps to get_devices, get_presentations, assign_schedule and a follow-up get_weekly_schedule to confirm the result. If two device groups match “lobby”, the assistant asks which one before it drafts the write.

How does MCP work with digital signage?

MCP, the Model Context Protocol, is the open standard Anthropic released in November 2024 that lets an AI assistant discover and call tools on an external system. A signage vendor runs an MCP server that publishes its operations as tools with typed inputs and safety annotations. The assistant connects over HTTPS, authenticates with OAuth, lists the tools, and calls them as JSON-RPC requests. The CMS does the work and the assistant handles the conversation.

The CastHub server at webapi.cast-hub.com/mcp uses OAuth 2.0 with PKCE and dynamic client registration. Each CastHub tool carries a ReadOnly, Destructive or Idempotent annotation, which tells Claude or ChatGPT when to ask before acting. How CastHub MCP works shows the full request flow, and the CastHub MCP overview lists the 8 tool domains.

Can you update digital signage from a Slack message?

Yes, in two different ways. OptiSigns, ScreenCloud and Valotalive publish Slack apps that mirror messages from chosen channels onto screens, so a post to a signage channel appears on the wall at the next refresh. That is a display feed, not control: it cannot change a schedule or swap a presentation. The assistant path is the second way, where you tell Claude or ChatGPT what to change and the assistant calls the CMS directly.

The Slack apps fit when the goal is announcements for deskless staff; the Slack to screens guide compares them. OptiSigns filters a wall to all messages, pinned messages or messages that mention @OptiSigns. Valotalive follows up to four channels per app. Beyond mirroring a feed, the CastHub MCP server lets an assistant publish a presentation or start a text alert on a named device group from any chat window.

What can be automated and what still needs a person?

An assistant can automate any change that is a defined CMS operation: schedule swaps, presentation assignment, alerts, device restarts, group membership and status checks. A person still writes the message, decides whether a closure or alert is warranted, approves the preview, and handles design work and anything the CMS has no tool for. CastHub does not sync prices from a point of sale, so a price change is typed once by a person and propagated by the assistant.

Hours change across 30 stores. Before, a coordinator opens each store’s schedule, edits the closing time and saves, about 90 minutes. After, one sentence to Claude naming the group and the new hours, a preview listing 30 devices, one confirmation, about 5 minutes.

Weather closure. Before, a manager on a phone at 5:30 AM logs in, finds the closure slide, assigns it to the entrance screens and must remember to revert it, 20 to 30 minutes. After, “show the closure notice on all entrance screens until 7 AM tomorrow”, preview, confirm, about 2 minutes, with the revert scheduled in the same request.

Price update. Before, half a day of per-store edits on a 23-store menu board. After, the new prices pasted into one request, a preview of the changed fields, confirm, under 5 minutes.

Emergency alert. Before, 5 to 10 minutes to find the alert asset and target the right screens. After, “start a text alert on the warehouse group: evacuate by the north exit”, preview, confirm, under a minute. The person decides the alert is needed; the assistant does the targeting.

The before figures are typical dashboard times CastHub hears from multi-location operators; the after figures include preview and confirmation. If the 30 stores are not yet a device group, the assistant creates the group in the same conversation.

What does AI-assisted signage automation cost?

AI-assisted signage automation costs the CMS plan plus any AI add-on the vendor charges. CastHub charges no AI add-on fee: the MCP server is included on every plan, at Basic $12 per month for up to 2 screens, Pro $35 per month for up to 10 screens and Enterprise $95 per month for up to 25 screens, with an unlimited-screen add-on at $79 per month. Yodeck includes MCP on Premium and Enterprise plans at no extra charge.

Revel Digital includes AI on every plan with token limits per managed player. The Amazon Signage Stick is $99.99 on Amazon, and CastHub runs on Fire TV, Apple TV, Android TV, Samsung Tizen, LG webOS and Windows hardware you own. Custom connector support in ChatGPT sits on its paid plans. Screenly, NoviSign, Skoop, ScreenCloud and OptiSigns did not publish AI pricing on the pages fetched on September 1, 2026. For per-screen versus flat-rate math, see the digital signage pricing guide.

Is AI control safe for live screens?

AI control is safe for live screens when the CMS gates every write behind a preview and a confirmation and separates read tools from write tools. CastHub does this with ReadOnly, Destructive and Idempotent annotations on each tool, so Claude or ChatGPT runs reads freely and asks before any change reaches a screen, over OAuth 2.0 with PKCE. Audit logging and rollback are on the CastHub roadmap and are not shipped, and scoped per-tool permissions are in development.

Yodeck disables delete operations through its AI agent and scopes the agent to your existing permissions. Screenly’s server inherits the permissions of the API token that configured it. Fugo puts a confirmation step on Actions. Revel Digital describes agent actions as opt-in and auditable and cites ISO 27001. Start with read-only requests, test writes on one test device group, and keep the confirmation prompt on; skipping the preview saves seconds and costs the check that matters.

Sources

Competitor facts were read from the pages below on September 1, 2026. Re-check before you buy.

Page	URL	Fetched
Revel Digital	www.reveldigital.com/ai-features	2026-09-01
NoviSign	www.novisign.com/software/ai/	2026-09-01
Fugo	www.fugo.ai/features/ai-digital-signage/	2026-09-01
Skoop	www.skoopsignage.com/features/skoop-ai	2026-09-01
Screenly	www.screenly.io/developers/mcp-server/	2026-09-01
Yodeck	www.yodeck.com/ai-agent/	2026-09-01
ScreenCloud AI content	screencloud.com/digital-signage/ai-digital-signage-content	2026-09-01
ScreenCloud Slack	screencloud.com/apps/slack	2026-09-01
OptiSigns Slack	support.optisigns.com/hc/en-us/articles/4407335862931-How-to-use-Slack-App	2026-09-01
Valotalive Slack	valota.live/apps/slack/	2026-09-01

To try assistant control on your own screens, the CastHub MCP overview shows the connector URL and tool list, and CastHub pricing starts at $12 per month with a 14-day trial, nothing charged until it ends.