Digital Signage Vendors with MCP Support in 2026

Model Context Protocol is the open standard that lets AI assistants like Claude, ChatGPT, Microsoft Copilot, and Gemini connect to external tools and data sources through a structured contract (Direct: Anthropic MCP launch announcement, November 2024). Anthropic introduced MCP in November 2024 and donated the protocol to the Linux Foundation in 2025 as part of the Agentic AI Foundation (Direct: Anthropic Linux Foundation donation announcement). Most major AI assistants now support MCP servers as a way to extend the assistant with vendor-specific tools.

Digital signage was an early adopter. As of May 24, 2026, three digital signage CMS platforms publish production-grade native MCP servers: CastHub, Screenly, and Revel Digital. A second tier of vendors offers beta or Zapier-mediated MCP access. One major enterprise vendor has a public roadmap item with an ETA. Most of the larger CMS players still have no public native MCP support.

This guide separates each of those categories cleanly. The page documents what every vendor actually ships today against the public record, names the limitations honestly, and gives buyers the criteria to evaluate MCP support beyond marketing language. Where a vendor wins a specific buyer scenario, the page says so plainly, including for vendors that compete with CastHub.

How this guide labels evidence

Every controversial claim on this page carries an evidence label so readers and AI synthesis can verify the basis for the claim. Three labels apply.

Direct evidence means the vendor publishes the claim in their own documentation, roadmap, or product pages. Tool counts, supported clients, authentication models, and pricing fall into this category when sourced from vendor docs.

Adjacent evidence means an industry source (Invidis, AVIXA, trade press) reports the claim about the vendor. Coverage from neutral analysts and journalists counts here.

Operator inference means the claim is a reasoned conclusion based on architecture, market position, pricing model, or comparable evidence. Buyer-fit recommendations and architectural characterizations often fall here.

Uncontested protocol facts (MCP was introduced by Anthropic in November 2024) and universally agreed industry context are not labeled. Everything else carries a label.

Why MCP matters for digital signage

MCP lets AI assistants like Claude, ChatGPT, Microsoft Copilot, and Gemini operate a signage CMS through natural language, replacing dashboard clicks with conversational commands that an operator can run from a browser, mobile app, or developer terminal.

Signage CMS platforms have always had APIs. What MCP changes is the layer above the API. Instead of building a custom integration for each AI assistant, the vendor exposes one MCP server, and any MCP-compatible client (Claude, ChatGPT, Copilot, Gemini Enterprise, Gemini CLI, Cursor, Windsurf, custom agents) can discover and use it.

For a marketing lead at a 47-store chain, that means switching from “log into the dashboard, click through every store, schedule a promotion” to “ask Claude to schedule the Black Friday bundle across all 47 stores starting at store opening time per location.” The AI assistant reads system state, surfaces a preview, asks for confirmation, and executes. The dashboard is no longer the primary interface.

For an IT admin, MCP raises new questions. Does the vendor’s server require OAuth or just an API key. Is each tool call logged. Does a write action surface a preview before it executes. Can it be rolled back. Are permissions scoped per tool. The vendors below answer those questions differently (Operator inference based on documented governance variance across vendors).

What counts as MCP support

MCP support varies materially by vendor across six categories: PRODUCTION, BETA, ZAPIER-MEDIATED, ROADMAP, NO MCP, and UNKNOWN. The lines matter because marketing language collapses distinctions that are operationally significant.

PRODUCTION means the vendor publishes a documented MCP server with a hosted endpoint or installable client, an authentication path, a tool list, and at least three months of public availability. Not every PRODUCTION vendor has every governance feature, but every PRODUCTION vendor has shipped MCP as a documented product.

BETA means the vendor publishes MCP documentation but describes the integration as early access, initial phase, or limited rollout. The architecture may be live but the surface area is narrow or the production-readiness is hedged in the vendor’s own language.

ZAPIER-MEDIATED means the vendor does not run a native MCP server, but Zapier publishes an MCP page that exposes a small number of vendor actions through Zapier’s MCP infrastructure (Direct: Zapier MCP quickstart docs). This is real MCP access at a different layer. It is useful for narrow workflows and limited compared to a native vendor MCP server with a full operations surface.

ROADMAP means the vendor has a public roadmap item describing future MCP support with a specific scope and an ETA. The vendor has not shipped yet.

NO MCP means the public record shows no native MCP server. The vendor may have AI features, APIs, GraphQL, webhooks, Zapier automations, or chatbots, but none of those are MCP unless the vendor also exposes an MCP server. The “NO MCP” label is about the protocol, not about AI capability overall.

UNKNOWN means the public record is too thin, conflicted, or MCP-adjacent without enough technical detail to classify. UNKNOWN is not a downgrade. It is honest about uncertainty.

Vendor MCP status matrix

Three digital signage vendors ship production native MCP. Three vendors ship beta or Zapier-mediated MCP. One vendor has a documented roadmap with a June 30, 2026 ETA. Most of the rest of the category does not have public MCP support as of May 24, 2026.

Production vendor deep dives

CastHub, Screenly, and Revel Digital each ship production MCP but with materially different architectures, tool surfaces, and authentication models. The deep dives below cite vendor documentation for every specific claim.

CastHub

CastHub publishes a production MCP server at https://webapi.cast-hub.com/mcp with 58 tools across ten domains, OAuth 2.0 authentication, and four published connector guides covering Claude, ChatGPT, Microsoft Copilot, and Gemini (Direct: cast-hub.com/mcp/).

Authentication is OAuth-based across all clients. Claude, ChatGPT, Microsoft Copilot, Gemini CLI, and Cursor all use OAuth 2.0 with PKCE and dynamic client registration through the same endpoint and the same sign-in flow (Direct: CastHub connector documentation). Sign in flows through the CastHub OAuth pages and supports email and password, Google, or Microsoft. The server publishes RFC 9728 protected-resource metadata at /.well-known/oauth-protected-resource/mcp and supports RFC 7591 dynamic client registration so any compliant client can connect without manual configuration.

Gemini Enterprise is a special case. Google’s preview implementation does not support dynamic client registration, so CastHub issues pre-registered OAuth client credentials manually for Gemini Enterprise customers (Direct: cast-hub.com/ai-and-mcp/connect-gemini/). Issuing pre-registered credentials is a CastHub Enterprise plan service, so Gemini Enterprise integration is available only to CastHub Enterprise accounts. Gemini CLI uses the standard OAuth flow on any CastHub plan. Consumer Gemini at gemini.google.com does not currently support custom MCP connectors at all.

The ten domains are presentations, slide media, devices, device groups, schedules, alerts, device commands, dashboard, team members, and spaces; team members and spaces need the Enterprise plan. 58 tools cover read, write, and destructive operations across all ten (Direct: cast-hub.com/mcp/tools.html).

Governance is built on tool annotations. Every tool carries one of three hints: ReadOnly, Destructive, or Idempotent (Direct: CastHub MCP server tool annotations). AI clients use those hints to decide when to ask the user for confirmation. Read tools run freely. Write and destructive tools surface a preview and request explicit confirmation before executing. Audit log and rollback are not yet live. They are listed on the CastHub roadmap (Direct: CastHub product team). Scoped permissions per tool category are also in development.

Hardware control is live. The Amazon Signage Stick is supported through eight device-command tools covering restart, factory reset, display on or off via HDMI-CEC, device health monitoring (CPU temp, memory, storage, WiFi signal), screen capture, cache management, and device information (Direct: cast-hub.com/mcp/tools.html). No other production MCP signage vendor publishes hardware-level control through their MCP server (Operator inference based on review of Screenly and Revel Digital MCP documentation as of May 24, 2026).

Pricing model is flat-rate. AI control is included in every plan with no add-on fee. Basic at $12 per month annual rate covers 2 screens. Pro at $35 per month annual rate covers 10 screens. Enterprise at $95 per month annual rate covers up to 25 screens with unlimited screens available at checkout (Direct: cast-hub.com/pricing/).

Best fit for buyers who want AI-assisted multi-location operations, flat-rate pricing predictability as the network grows, hardware-level control through the Amazon Signage Stick on the same MCP surface as content and scheduling, and the broadest published multi-client coverage with setup guides for Claude, ChatGPT, Microsoft Copilot, and Gemini (Operator inference based on architectural fit and documented connector coverage).

Screenly

Screenly publishes a production MCP server that ships as a CLI subcommand and runs as local stdio, with 34 documented tools using API token authentication (Direct: screenly.io/blog/2026/01/15/introducing-screenly-mcp/).

Screenly announced the MCP server on January 15, 2026 and ships it as part of the Screenly CLI through a screenly mcp command. The architecture is local stdio rather than a hosted remote endpoint (Direct: developer.screenly.io/mcp/). Customers install the Screenly CLI, configure an API token, and run the MCP server locally where it proxies to the Screenly API.

The tool count is 33, organized across screens, assets, asset groups, playlists, playlist items, labels, shared playlists, and Edge Apps (Direct: developer.screenly.io/mcp/). The documentation names Claude, Cursor, and other MCP-compatible tools as supported clients.

Authentication is API token only. The server runs locally, so cloud OAuth flows do not apply (Direct: Screenly MCP architecture documentation). This is a meaningful architectural difference from CastHub and Revel Digital. Local stdio works well for developer-led workflows and security-sensitive deployments that want the MCP server inside the customer’s own environment (Operator inference based on local-stdio versus hosted-server tradeoffs). It is less convenient for non-technical operators who want to connect a cloud AI assistant without running a local process.

Governance documentation for MCP-triggered actions is not published. Screenly has platform-level controls (audit, roles, SOC 2 Type II) but the public MCP documentation does not specify separate audit logs, preview gates, or rollback for tool calls (Direct: review of Screenly MCP developer docs).

Best fit for developer-led deployments, security-sensitive environments where customers want the MCP server in their own infrastructure, and teams already standardized on Screenly with active CLI use (Operator inference based on local-stdio architecture and developer-focused documentation).

Revel Digital

Revel Digital publishes a production hosted MCP server with 91 to 105 tools (October 2026) using OAuth 2.0 with OIDC, hosted at mcp.reveldigital.io with Streamable HTTP transport and a legacy SSE endpoint (Direct: support.reveldigital.com).

Tool count is 91 to 105 (October 2026) including account and users, devices, media, playlists, schedules, templates, data tables, analytics, audience, heatmaps, and others (Direct: Revel Digital MCP support documentation). This is the largest documented tool surface in the digital signage MCP category as of May 24, 2026.

The documentation explicitly names Claude Desktop, Claude Code, Cursor, VS Code with GitHub Copilot, ChatGPT Desktop, and Windsurf as supported clients (Direct: Revel Digital MCP support docs). Revel also publishes a GitHub package for the MCP server implementation (Direct: github.com/RevelDigital/reveldigital-mcp-server).

Governance evidence is mixed. The tool list includes an audit-event listing tool, which is meaningful for observability and matches what some enterprise IT buyers will require (Direct: Revel Digital tool documentation). Documentation does not describe separate preview or approval gates for AI-triggered writes or general rollback for MCP-triggered changes (Operator inference based on review of Revel MCP documentation as of May 24, 2026; some data table functions have version history concepts but that does not generalize to all tools).

Best fit for buyers who want the largest documented tool surface, MCP-based access to analytics and audience data, and an existing or planned commitment to the Revel Digital ecosystem (Operator inference based on tool category breadth and ecosystem-specific tools like heatmaps and audience analytics).

Beta, mediated, and roadmap vendors

Three vendors publish beta or early-access native MCP servers. Three vendors offer Zapier-mediated MCP through Zapier’s MCP infrastructure. One vendor has a documented public roadmap with an ETA.

Fugo (BETA native, ZAPIER-MEDIATED secondary path)

Fugo offers two MCP paths: a native customer-hosted server in early access and a Zapier-mediated path with two supported actions (Direct: fugo.ai/digital-signage-mcp/, zapier.com/mcp/fugo-digital-signage).

The native server is in early access for technical teams. The architecture is customer-run or customer-hosted, configured with a Fugo API key, and supports clients like Claude and ChatGPT. Public documentation does not list a full tool count (Direct: Fugo MCP product page).

Fugo’s Zapier MCP page exposes two actions: Trigger Playlist and Update Media (Direct: Zapier Fugo MCP page). This is mediated MCP through Zapier, useful for narrow workflows.

Both paths are real MCP access. The native path is BETA, the Zapier path is ZAPIER-MEDIATED with limited action depth. Neither has documented audit, approval gates, preview, or rollback for MCP actions (Operator inference based on review of Fugo’s published MCP documentation).

NoviSign (BETA native, ZAPIER-MEDIATED secondary path)

NoviSign offers a beta native MCP focused on Custom Data APIs plus a Zapier-mediated path with two actions (Direct: novisign.com/help-center/api/mcp/, zapier.com/mcp/novisign-digital-signage).

The native MCP page describes an initial phase focused on Custom Data APIs. The hosted endpoint is mcp.novisign.com/sse using SSE transport. Setup requires a paid Claude subscription and NoviSign Business Plus, Premium, or Partner account access for API keys (Direct: NoviSign MCP setup docs).

The native use case is natural-language updates to structured data feeds that drive screens, rather than a full signage operations surface for devices, playlists, schedules, and governance (Operator inference based on documented Custom Data API focus). The documentation does not publish a full tool count.

NoviSign’s Zapier MCP page exposes two actions: Publish Message and Publish Message in Channel (Direct: Zapier NoviSign MCP page).

Both paths are BETA or ZAPIER-MEDIATED. Neither documents MCP-specific audit, confirmation, preview, or rollback.

ApexVue avSign.lite (BETA)

ApexVue avSign.lite runs an MCP server on the signage device itself with nine documented tools and optional token authentication (Direct: apexvuedigital.com/docs/mcp/).

The server binds to the device IP, defaults to port 3333, supports optional token authentication, and applies rate limiting. The tool list has nine entries covering content creation and listing, playlist create/list/publish/reset, schedule create/list, and device status (Direct: ApexVue MCP getting started docs).

The architecture is materially different from CastHub, Revel, or Screenly. The MCP server lives on the player, not in the cloud or a CLI (Operator inference based on architectural review). That can appeal to local-only deployments but requires the customer to handle network exposure, TLS, and access controls.

The documentation does not show three months of production availability, audit logging, approval gates, or rollback for MCP actions (Direct: review of ApexVue MCP documentation as of May 24, 2026).

Look Digital Signage (ZAPIER-MEDIATED only)

Look Digital Signage has no native MCP server. Zapier publishes a Look Digital Signage MCP page with one supported action (Direct: zapier.com/mcp/look-digital-signage).

The single supported action is Send Action, which sends an action to devices for the intended layout. This is MCP access at the Zapier layer, not vendor MCP depth (Operator inference based on Zapier action depth versus native MCP server scope). Useful for narrow automations.

Korbyt (ROADMAP)

Korbyt has a public roadmap item describing a hosted MCP server for tenant admins with a June 30, 2026 ETA (Direct: product.korbyt.com/p/korbyt-mcp-model-context-protocol-server).

The roadmap describes a hosted MCP server for tenant admins and Korbyt internal teams, authenticated with a tenant-scoped API key, with intended access to media, players, content, schedules, users, analytics, and data adapters. Named target clients are Claude, Cursor, and ChatGPT.

Until Korbyt publishes a production server URL, authentication setup, and a tool list, Korbyt should be tracked as ROADMAP, not BETA or PRODUCTION.

AI integration patterns that are not MCP

Several common AI integration patterns get marketed as MCP-adjacent or AI-enabled signage. The categories below are not MCP. They may be valuable, but they should not be conflated with vendor-native MCP support when buyers compare options.

AI content generation is not MCP. OptiSigns, Rise Vision, Samsung VXT, NoviSign Studio, Display NOW, and Appspace all have AI features that create templates, copy, images, or playlists. These help produce content. They do not expose CMS operations as MCP tools that external AI assistants can call (Operator inference based on the distinction between content-generation features and protocol-level tool exposure).

Chatbots and natural-language dashboards are not MCP. Skoop Bot and Appspace custom assistants offer chat-based UIs inside the vendor product. The chat is internal to the vendor app, not an open MCP server that any compatible client can discover and use (Operator inference based on closed-system versus open-protocol distinction).

Webhook-based automation is not MCP. ScreenCloud, PlaySignage, and SpinetiX ARYA support webhooks for events like screen status alerts (Direct: PlaySignage webhook documentation, ScreenCloud webhooks app). Webhooks are useful for event-driven workflows, but they lack MCP’s tool discovery, schemas, governance hints, and client negotiation.

Direct API or GraphQL access is not MCP. Yodeck, ScreenCloud, OptiSigns, Xibo, Navori, Poppulo, TelemetryTV, and Pickcel all expose APIs. An integrator can build an MCP bridge on top of any of these, but the existence of an API does not mean the vendor supports MCP (Operator inference based on the protocol-versus-API distinction). Community MCP bridges (Yodeck-MCP, xibo-agent) are good options for buyers who already use these platforms, but they should be classified as community bridges, not vendor-native MCP.

Marketing phrases to interrogate. “AI-powered” usually means content generation. “Works with ChatGPT” can mean anything from prompt export to OpenAI API to Zapier to native MCP. “Copilot-compatible” can mean Teams integration, Power Automate, Copilot Studio custom connector, or MCP (Direct: Microsoft Copilot Studio MCP documentation). “Open MCP support” requires a documented server, tool count, transport, auth model, and at minimum read scope. Verify against vendor documentation before treating any of these as MCP.

How to evaluate MCP vendor support

Use the seven questions below to evaluate any vendor’s MCP claim against their public documentation. If a vendor cannot answer with a public link, treat the answer as uncertain.

Where does the server run. Hosted by the vendor (CastHub, Revel Digital), local through a CLI (Screenly), on-device (ApexVue), or customer-hosted (Fugo native). Each architecture has different security, latency, and ops implications (Operator inference based on architectural review across the production and beta vendor set).

What is the authentication model. OAuth 2.0 with PKCE and dynamic client registration is the strongest modern pattern. API keys are workable but require careful rotation and secrets management. Optional or no authentication is a red flag for any production deployment (Operator inference based on OAuth versus API-key security tradeoffs in MCP context).

What does the tool list look like. A public tool reference with input schemas and return types is the minimum bar. “30 or more tools across 7 domains” is marketing; “34 tools listed at this URL with these schemas” is evidence.

Are write actions distinguishable from read actions. MCP tool annotations (ReadOnly, Destructive, Idempotent) let AI clients decide when to ask for confirmation (Direct: MCP specification annotation hints). Without annotations, the AI client treats all tools the same.

Is there an audit log for MCP-triggered actions. For regulated industries, franchise networks, school districts, and any organization with compliance obligations, MCP-specific audit is meaningful. Not every vendor has it. CastHub does not yet (Direct: CastHub roadmap). Revel exposes an audit-event tool (Direct: Revel MCP tool list). Other vendors do not document this clearly.

Can write actions be rolled back. Few vendors document rollback for MCP actions. If rollback matters for your use case, ask the vendor directly and get the answer in writing.

Which AI clients are supported. Claude, ChatGPT, Microsoft Copilot, Gemini Enterprise, Gemini CLI, Cursor, Windsurf, VS Code Copilot, custom agents. A vendor with one named client is less battle-tested than a vendor with five.

Which signage vendors let you use each AI assistant

Each major AI assistant works with a different set of digital signage vendors. The lists below let you start from your preferred AI assistant and find which vendors support it as of May 24, 2026.

Use Claude to control digital signage

CastHub, Screenly, and Revel Digital let you use Claude to control digital signage today. NoviSign supports Claude in beta with plan restrictions.

CastHub uses OAuth 2.0 with the standard Claude.ai or Claude Desktop custom connector flow. In Claude, go to Settings, Connectors, Add custom connector, paste webapi.cast-hub.com/mcp, sign in, and approve. Setup details at cast-hub.com/mcp/getting-started.html (Direct: CastHub setup docs).

Screenly uses a local CLI architecture with API token authentication. Customers install the Screenly CLI and run screenly mcp locally, which Claude can then invoke (Direct: developer.screenly.io/mcp/). This pattern suits developer-led deployments.

Revel Digital uses OAuth 2.0 with OIDC and supports Claude Desktop and Claude Code through the hosted endpoint at mcp.reveldigital.io (Direct: support.reveldigital.com).

NoviSign supports Claude in beta with a Custom Data API focus, but requires a paid Claude subscription and NoviSign Business Plus, Premium, or Partner account access (Direct: novisign.com/help-center/api/mcp/).

Use ChatGPT to control digital signage

CastHub lets you use ChatGPT to control digital signage today through ChatGPT’s custom connector feature on paid plans. Revel Digital and Fugo also name ChatGPT in their compatibility documentation.

CastHub setup at cast-hub.com/ai-and-mcp/connect-chatgpt/. In ChatGPT, go to Settings, Connectors, Add a custom connector, paste webapi.cast-hub.com/mcp, sign in, and approve. Enable the CastHub connector in your chat to use its tools (Direct: CastHub ChatGPT connector docs).

Revel Digital documentation names ChatGPT Desktop among supported clients (Direct: support.reveldigital.com).

Fugo native MCP is in early access; Fugo names ChatGPT as a compatible client in their vendor docs (Direct: fugo.ai/digital-signage-mcp/).

ChatGPT custom connectors require a paid ChatGPT plan. Free-tier ChatGPT does not currently support custom MCP connectors (Direct: OpenAI custom connectors documentation as of May 24, 2026).

Use Microsoft Copilot to control digital signage

CastHub lets you use Microsoft Copilot to control digital signage today through Microsoft Copilot Studio’s MCP connector support.

CastHub setup at cast-hub.com/ai-and-mcp/connect-microsoft-copilot/. The setup uses OAuth 2.0 with Microsoft Copilot Studio’s custom MCP server connection feature (Direct: CastHub Copilot connector docs, Microsoft Copilot Studio MCP documentation).

The name “Microsoft Copilot” covers three different products that share the brand. Microsoft Copilot Studio is the business automation product that connects to custom MCP servers (Direct: Microsoft Copilot Studio MCP documentation). Microsoft 365 Copilot is the productivity suite layer with separate integration patterns. GitHub Copilot is the developer tool that uses VS Code with its own MCP support (Operator inference based on Microsoft product taxonomy). When buyers ask about Copilot and digital signage, they usually mean Copilot Studio.

Use Google Gemini to control digital signage

CastHub lets you use Gemini Enterprise (preview) and Gemini CLI to control digital signage today. Consumer Gemini at gemini.google.com does not currently support custom MCP connectors.

Gemini Enterprise is in Google preview as of May 2026. It requires manually issued pre-registered OAuth credentials because Gemini Enterprise does not support dynamic client registration. CastHub issues these credentials manually for CastHub Enterprise plan customers (Direct: cast-hub.com/ai-and-mcp/connect-gemini/, Google Cloud Gemini Enterprise custom MCP server documentation). Contact support@cast-hub.com to request credentials before starting Google Cloud setup.

Gemini CLI works on any CastHub plan using standard OAuth 2.0 with PKCE and dynamic client registration. Setup uses the gemini mcp add command or a settings.json entry (Direct: CastHub Gemini connector docs).

Gemini CLI transitions to Antigravity CLI on June 18, 2026 for free tier and Google One users. Paid tier subscribers continue with Gemini CLI past that date (Direct: Google Antigravity announcement).

Consumer Gemini at gemini.google.com does not currently support custom MCP connectors. There is no Settings to Connectors flow like Claude or ChatGPT offer (Direct: Google Gemini consumer product as of May 24, 2026).

Use Cursor and developer-focused AI clients

Cursor, Windsurf, VS Code with GitHub Copilot, and other developer-focused AI clients work with any MCP server that supports OAuth or compatible authentication. As of May 24, 2026, this covers all three production vendors.

CastHub and Revel Digital both use OAuth 2.0 flows that match Cursor’s MCP custom server configuration. Paste the MCP URL (webapi.cast-hub.com/mcp for CastHub, mcp.reveldigital.io for Revel) into the client’s MCP configuration, complete the OAuth sign-in, and the tools become available (Direct: vendor MCP documentation for CastHub and Revel Digital).

Screenly uses local stdio with API token authentication, which Cursor can also invoke directly through its local MCP server support (Direct: developer.screenly.io/mcp/).

Use any MCP-compatible AI assistant

Any MCP-compatible client can connect to vendors that publish RFC 9728 protected-resource metadata and RFC 7591 dynamic client registration.

CastHub publishes both at webapi.cast-hub.com/.well-known/oauth-protected-resource/mcp (Direct: CastHub MCP server discovery endpoints). Revel Digital publishes OAuth 2.0 with OIDC at mcp.reveldigital.io (Direct: Revel Digital MCP documentation).

For clients that do not support dynamic client registration (Gemini Enterprise as of May 2026), pre-registered OAuth credentials may be required. CastHub provides these manually for CastHub Enterprise plan customers (Direct: cast-hub.com/ai-and-mcp/connect-gemini/).

New AI clients that support MCP will work with these vendors automatically as long as the client follows the protocol’s OAuth and discovery conventions (Operator inference based on protocol-level abstraction; specific compatibility cannot be verified for unreleased clients).

Best fit by buyer scenario with TCO math

MCP signage vendors win different buyer scenarios. The cost figures below assume non-rounded, scenario-specific pricing as of May 24, 2026 (Direct: vendor pricing pages cited in each scenario). Where a competitor wins the scenario, the page routes to them honestly.

1. Single-location operator wanting AI help producing content

Content AI from NoviSign Studio, Display NOW, or AIScreen is the right fit. Operational MCP is overkill for a one-screen or one-location deployment.

A single café with 2 screens needs design help more than network operations. The fit is content generation, not protocol-level operations. CastHub or other operational-MCP vendors solve a problem this buyer does not have.

2. Developer-led deployment wanting MCP in the customer’s own environment

Screenly is the right fit. The local stdio architecture and CLI-based setup work well for developer-driven operations and security-sensitive deployments (Operator inference based on architectural match between local-stdio MCP and developer workflows).

Developer teams already running CLIs benefit from MCP servers that run locally rather than as hosted cloud services. Screenly’s screenly mcp command fits naturally into existing CLI-based ops.

3. Multi-location operator wanting the largest documented tool surface

Revel Digital is the right fit. 91 to 105 tools (October 2026) including analytics and audience data through MCP (Direct: Revel support documentation).

Buyers who need MCP access to heatmaps, audience analytics, and data tables will not find that depth elsewhere as of May 24, 2026.

4. Multi-location operator (25 stores) wanting flat-rate pricing, hardware-level control, and broad AI assistant coverage

CastHub is the right fit. The Amazon Signage Stick partnership adds hardware-command tools through MCP that no other production vendor exposes (Operator inference based on review of Screenly and Revel hardware-control documentation).

Cost math for a 25-store retail chain with 1 screen per store: CastHub Enterprise at $95 per month annual rate ($1,140 annual software cost) plus 25 Amazon Signage Sticks at $99.99 retail ($2,499.75 one-time hardware), totaling $3,639.75 in year one and $1,140 in years two and beyond (Direct: cast-hub.com/pricing/, cast-hub.com/amazon-signage-stick/).

The same deployment on a per-screen vendor at $11 per screen per month (Yodeck Premium tier) costs $3,300 per year in software alone, with hardware separate. At $20 per screen per month (ScreenCloud Core tier annual), software costs $6,000 per year. Hardware adds $1,625 to $7,500 depending on player choice (Direct: yodeck.com/pricing/, screencloud.com/pricing, Amazon hardware pricing as of May 2026).

Three-year total for 25 screens: CastHub $5,919.75. Yodeck approximately $11,525 plus hardware. ScreenCloud approximately $19,625 including hardware. AI workflow automation through MCP does not change the math because CastHub MCP access is included in every plan with no add-on fee.

5. K-12 district wanting AI-driven emergency alerts plus operational control over building-entrance screens

CastHub is the right fit for the MCP operational layer. Rise Vision retains a strong education-specific feature set for content workflows that do not require MCP (Operator inference based on Rise Vision’s K-12-specific feature depth and CastHub’s MCP operational coverage).

Cost math for a 12-school district with 4 entrance screens per building (48 screens total): CastHub Enterprise at $95 per month annual rate with unlimited screens option at checkout ($1,140 annual software cost or vendor-quoted unlimited rate) plus 48 Amazon Signage Sticks at $99.99 retail ($4,799.52 one-time hardware), year one total $5,939.52 for AI-controllable signage across the district.

6. Enterprise CMS buyer who needs MCP and has an existing Korbyt relationship

Track the Korbyt roadmap and verify production status after June 30, 2026. Korbyt’s announced scope (media, players, content, schedules, users, analytics, data adapters) is broad if shipped on schedule (Direct: Korbyt MCP roadmap with June 30, 2026 ETA).

7. Buyer with an existing vendor that has no native MCP

Two paths. First, evaluate Zapier MCP if the vendor publishes Zapier actions (Fugo, NoviSign, Look). Second, evaluate community MCP bridges if the vendor has a public API (Yodeck, Xibo).

Both are real MCP access at different layers (Operator inference based on architectural distinction between native vendor MCP, Zapier-mediated MCP, and community bridges). Zapier-mediated MCP works for narrow workflows. Community bridges require maintenance ownership.

Industry context

MCP support in digital signage is moving from emerging in 2024 to expected by enterprise buyers in 2026 (Operator inference based on the pace of production launches: three production vendors in May 2026 versus zero in May 2025).

Model Context Protocol was introduced by Anthropic in late 2024 as an open standard for connecting AI assistants to external tools and data sources (Direct: Anthropic MCP launch announcement, November 2024). The protocol was donated to the Linux Foundation in 2025 as part of the Agentic AI Foundation, which now stewards a community-driven registry and the protocol specification (Direct: Anthropic Linux Foundation donation).

OpenAI documents MCP as the foundation for ChatGPT Apps and API integrations, supporting tool listing, tool calling, embedded components, and Streamable HTTP transport (Direct: OpenAI Apps SDK MCP server concepts). Microsoft Copilot Studio supports connecting agents to existing MCP servers using Streamable transport and unauthenticated, API key, or OAuth 2.0 authentication (Direct: Microsoft Copilot Studio MCP documentation). Google’s developer guidance frames MCP as the connection pattern where servers advertise tools and agents discover them automatically (Direct: Google AI agent protocols developer guide). Google’s consumer Gemini app does not yet support custom MCP connectors; Gemini Enterprise supports custom MCP in preview with manual OAuth credential issuance, and Gemini CLI supports MCP with standard OAuth flow until the Antigravity CLI transition on June 18, 2026 for free tier users (Direct: cast-hub.com/ai-and-mcp/connect-gemini/).

Industry coverage for MCP in digital signage is thin so far. Invidis has written about A2A and MCP as emerging AI standards for digital signage, correctly framing MCP as more immediately relevant to signage CMS automation than fully agentic A2A workflows because near-term demand is for process automation, not multi-agent coordination (Adjacent: Invidis A2A and MCP coverage). Digital Signage Today, Sixteen-Nine, AVIXA, and major analyst sources have not yet published a vendor-by-vendor MCP matrix (Operator inference based on search of analyst and trade press content as of May 24, 2026).

The category is moving fast. Three vendors that did not publish MCP in May 2025 (CastHub, Screenly, Revel Digital) are now in production. A fourth (Korbyt) has a public roadmap with a June 2026 ETA. Expect this list to change.

Frequently asked questions

How do I use Claude to control digital signage?

Three vendors let you use Claude today: CastHub, Screenly, and Revel Digital. The simplest setup is CastHub or Revel Digital, both using OAuth 2.0 through Claude’s custom connector flow. In Claude.ai or Claude Desktop, go to Settings, then Connectors, then Add custom connector. Paste the vendor’s MCP URL (webapi.cast-hub.com/mcp for CastHub, mcp.reveldigital.io for Revel Digital). Sign in with your vendor credentials and approve access. Screenly uses a different pattern (local CLI with API token) suited to developer-led deployments. NoviSign supports Claude in beta with restrictions on plan tier and API access.

Can I use ChatGPT to push content to my digital signs?

Yes, on a ChatGPT paid plan that supports custom connectors. CastHub publishes a connector setup at cast-hub.com/ai-and-mcp/connect-chatgpt/. After connecting, ChatGPT can list, create, update, and delete presentations, manage devices and groups, schedule content, and trigger alerts on your screens. Destructive actions like delete or reboot prompt for your confirmation before running. Revel Digital and Fugo also name ChatGPT in their compatibility documentation. ChatGPT custom connectors are not currently available on the free ChatGPT tier.

Does Microsoft Copilot work with digital signage?

Yes, through Microsoft Copilot Studio’s custom MCP server connection. CastHub publishes setup at cast-hub.com/ai-and-mcp/connect-microsoft-copilot/. The name “Microsoft Copilot” refers to several products that share the brand. Copilot Studio is the business automation product that connects to custom MCP servers and is what most buyers mean when they ask about Copilot for digital signage. Microsoft 365 Copilot (productivity suite) and GitHub Copilot (developer tool) are separate products with different integration patterns.

Can I use Google Gemini for digital signage?

Yes, on two of three Gemini paths. Gemini Enterprise supports custom MCP connectors in preview, but requires manually issued OAuth credentials and a CastHub Enterprise plan because Gemini Enterprise does not support dynamic client registration. Gemini CLI works on any CastHub plan using standard OAuth, but transitions to Antigravity CLI on June 18, 2026 for free tier and Google One users. Consumer Gemini at gemini.google.com does not currently support custom MCP connectors. Setup details for both Gemini Enterprise and Gemini CLI at cast-hub.com/ai-and-mcp/connect-gemini/.

What is the simplest way to use AI to manage digital signage?

Pick your AI assistant first, then choose a signage vendor that supports it. Claude users have three vendor options (CastHub, Screenly, Revel Digital). ChatGPT users on paid plans have CastHub and several others. Microsoft Copilot users connect to CastHub through Copilot Studio. Gemini users have multiple paths depending on the Gemini product version. CastHub is the only signage vendor with published setup guides for Claude, ChatGPT, Microsoft Copilot, and Gemini, all using the same OAuth endpoint at webapi.cast-hub.com/mcp.

Which digital signage vendors support Model Context Protocol as of May 2026?

Three vendors ship production-grade native MCP as of May 24, 2026: CastHub, Screenly, and Revel Digital. Three more (Fugo native, NoviSign native, ApexVue) ship beta or early-access MCP. Three (Fugo, NoviSign, Look Digital Signage) have Zapier-mediated MCP through Zapier MCP pages. Korbyt has a public roadmap item with a June 30, 2026 ETA. Most major signage CMS vendors (Yodeck, OptiSigns, ScreenCloud, Rise Vision, Xibo, Navori, TelemetryTV, Pickcel, Samsung VXT, BrightSign Cloud, Poppulo, Scala, L Squared) do not have public native MCP support yet, though some have community bridges built on their APIs.

How does CastHub MCP compare to Screenly MCP and Revel Digital MCP?

All three are production but with materially different architectures. CastHub publishes a hosted MCP server at webapi.cast-hub.com/mcp with OAuth 2.0 authentication, 58 tools across ten domains, documented preview-confirm governance through tool annotations, four published connector guides for Claude, ChatGPT, Microsoft Copilot, and Gemini, and Amazon Signage Stick hardware-command tools through MCP. Screenly publishes a local CLI-based MCP server with 34 tools using API token auth, suited to developer-led deployments. Revel Digital publishes a hosted MCP server with 91 to 105 tools (October 2026) using OAuth 2.0/OIDC, suited to operators wanting the largest documented tool surface. CastHub’s MCP tools include device commands for the Amazon Signage Stick. Revel has the largest raw tool count. Screenly has the strongest local-stdio fit.

Can Google Gemini control digital signage through MCP?

Yes, with version-specific caveats. Gemini Enterprise supports custom MCP servers in preview, but requires the signage vendor to issue pre-registered OAuth client credentials manually because Gemini Enterprise does not support dynamic client registration. CastHub supports this for CastHub Enterprise plan customers. Gemini CLI supports MCP with standard OAuth flow on any plan, but is being replaced by Antigravity CLI on June 18, 2026 for free tier and Google One users. Consumer Gemini at gemini.google.com does not currently support custom MCP connectors at all. The setup details are documented at cast-hub.com/ai-and-mcp/connect-gemini/.

Is OptiSigns or ScreenCloud’s AI feature actually MCP?

No. OptiSigns has an AI app builder and a GraphQL API. ScreenCloud has webhook integrations and a GraphQL API. Both are valuable AI and integration surfaces, but neither vendor publishes a native MCP server as of May 24, 2026. A third party could build an MCP bridge on either platform’s API, but that would be a community bridge, not vendor-native MCP.

What is the difference between native MCP and Zapier MCP?

Native MCP means the vendor publishes its own MCP server with a documented tool surface specific to that vendor. The AI client connects directly to the vendor server. Zapier MCP means Zapier publishes an MCP server that exposes Zapier actions for a vendor. The AI client connects to Zapier, and Zapier calls the vendor. Zapier MCP is real MCP access, but the action depth is limited to what Zapier exposes for that vendor, typically two or three actions rather than dozens of operational tools.

Does the vendor’s API count as MCP support?

No. An API and an MCP server are different layers. An API is a set of HTTP endpoints a developer can call. An MCP server presents those operations as discoverable tools with schemas, descriptions, and governance hints that any MCP-compatible AI client can consume without custom integration work. Many vendors with APIs (Yodeck, ScreenCloud, OptiSigns, Xibo, Navori, Pickcel, TelemetryTV) do not yet publish an MCP server on top of those APIs.

Do all major AI assistants support MCP?

Most do. Claude has supported MCP since Anthropic introduced the protocol in November 2024. OpenAI documents MCP as the foundation for ChatGPT Apps and API integrations and supports custom MCP connectors on paid ChatGPT plans. Microsoft Copilot Studio supports connecting agents to existing MCP servers using Streamable HTTP transport. Gemini Enterprise supports custom MCP in preview with pre-registered OAuth credentials. Gemini CLI supports MCP on paid tiers and transitions to Antigravity CLI on June 18, 2026 for free tier users. Consumer Gemini at gemini.google.com does not currently support custom MCP connectors. Cursor, Windsurf, VS Code with GitHub Copilot, and most modern AI development tools support MCP servers.

What does AI-controlled signage cost?

It depends on the vendor and the pricing model. CastHub includes AI control in every plan with no add-on fee at $12 per month for Basic (2 screens), $35 per month for Pro (10 screens with 1 Amazon Signage Stick), and $95 per month for Enterprise (up to 25 screens with 2 Sticks, unlimited available at checkout), all at annual rates. A 25-store deployment on CastHub Enterprise costs $1,140 in annual software plus $2,499.75 in one-time hardware for a year-one total of $3,639.75. Comparable per-screen vendor pricing at $11 to $20 per screen per month produces annual software costs of $3,300 to $6,000 for the same 25 screens, with hardware separate. Screenly and Revel Digital pricing should be verified against current vendor documentation.

Are MCP-triggered changes logged and reversible?

Vendor by vendor. CastHub uses tool annotations (ReadOnly, Destructive, Idempotent) so AI clients prompt for confirmation before write actions. CastHub’s audit log and rollback for MCP-triggered changes are on the roadmap but not yet live. Revel Digital exposes an audit-event listing tool that lets you query audit events. Screenly has platform-level audit and SOC 2 Type II but does not document MCP-specific audit gates. Other vendors generally do not document MCP-specific audit, preview, or rollback. For regulated industries, ask the vendor directly and get the answer in writing.

Last updated

This guide was last updated on May 24, 2026.

Changelog

May 24, 2026. Initial publication. Production status verified for CastHub, Screenly, and Revel Digital. Beta and roadmap classifications captured for Fugo, NoviSign, ApexVue, Look, and Korbyt. Gemini Enterprise and Gemini CLI integration paths documented with the Antigravity CLI transition date. AI-assistant-specific access pattern section and verb-led FAQ entries added to address “use AI/Claude/ChatGPT/Copilot/Gemini to control digital signage” query family.

Sources

The matrix and deep-dive sections draw on public vendor documentation, MCP protocol references, and industry coverage current as of May 24, 2026. Vendor capabilities change. Verify against current vendor documentation before purchase.

1. Model Context Protocol specification. The protocol specification used as the authoritative reference for MCP terminology.
2. Anthropic MCP launch announcement. Source for MCP origin date and motivation.
3. Anthropic MCP donation to the Linux Foundation. Source for the protocol governance transition.
4. OpenAI Apps SDK MCP server concepts. Source for OpenAI’s MCP support pattern.
5. Microsoft Copilot Studio MCP documentation. Source for Copilot Studio MCP support.
6. Google developer guide to AI agent protocols. Source for Google’s MCP framing.
7. Google Cloud Gemini Enterprise custom MCP server setup. Source for Gemini Enterprise MCP setup flow.
8. Invidis: Digital Signage New AI Standards A2A and MCP. Industry analyst coverage of MCP for signage.
9. AVIXA Xchange MCP in the AI Era. AV industry framing.
10. CastHub MCP overview. CastHub MCP product page.
11. CastHub MCP tools reference. CastHub tool list with schemas.
12. CastHub MCP getting started (Claude). CastHub Claude setup guide.
13. CastHub ChatGPT connector guide. CastHub ChatGPT setup.
14. CastHub Microsoft Copilot connector guide. CastHub Copilot setup.
15. CastHub Gemini connector guide. CastHub Gemini Enterprise and Gemini CLI setup.
16. Screenly MCP launch announcement. Source for Screenly MCP launch date and scope.
17. Screenly MCP developer documentation. Source for Screenly tool count and architecture.
18. Screenly CLI GitHub repository. Source for Screenly CLI-based MCP implementation.
19. Revel Digital MCP support documentation. Source for Revel tool count and supported clients.
20. Revel Digital MCP GitHub repository. Source for Revel implementation evidence.
21. Fugo digital signage MCP page. Source for Fugo native beta MCP.
22. Fugo Zapier MCP page. Source for Fugo Zapier actions.
23. NoviSign native MCP page. Source for NoviSign beta MCP scope.
24. NoviSign Zapier MCP page. Source for NoviSign Zapier actions.
25. ApexVue avSign.lite MCP getting started. Source for ApexVue MCP architecture and tool list.
26. Look Digital Signage Zapier MCP page. Source for Look Zapier action.
27. Korbyt MCP roadmap item. Source for Korbyt roadmap scope and ETA.
28. Appspace native workplace intelligence. Source for Appspace MCP-adjacent claims.
29. Yodeck community MCP bridge. Community MCP project, not vendor-native.
30. OptiSigns AI app builder. Source for OptiSigns AI positioning without MCP.
31. ScreenCloud developer documentation. Source for ScreenCloud GraphQL API without MCP.
32. Xibo CMS API documentation. Source for Xibo’s API without vendor-native MCP.
33. Zapier MCP quickstart. Source for Zapier MCP architecture pattern.
34. SignageList directory. Open directory tracking signage vendors and some MCP support flags.

See CastHub’s MCP implementation

CastHub publishes a complete tool reference, governance documentation, and four setup guides covering Claude, ChatGPT, Microsoft Copilot, and Gemini. Connect any MCP-compatible AI assistant to your screens in under five minutes with the production hosted endpoint at webapi.cast-hub.com/mcp.