1. General information

AudiencePlayer now offers an official MCP (Model Context Protocol) connector, being added to the Claude and ChatGPT connector/app directories (currently in review), and available today to any MCP-capable AI agent by URL. It lets an AI agent act as an authenticated admin user against your specific AudiencePlayer project, directly through natural-language conversation.

• You can find more information about general API-access here:
  ☞ https://support.audienceplayer.com/portal/en/kb/articles/audienceplayer-api-access
  
  

2. Connecting an AI agent

There are two ways to connect: through the connector directory of your AI product, or directly by URL. The AudiencePlayer MCP listings for Claude and ChatGPT are currently in review and will appear in their directories once approved; connecting by URL works today.

• Once the listing is approved: in Claude or ChatGPT, add the "AudiencePlayer MCP" connector from their respective app/connector directories. Before that, connect by URL as described below.
• You’ll be redirected through our normal SSO login. In addition to your usual admin login, you must explicitly grant a new permission scope,"API MCP Access" on top of your existing "Admin API access".
• Once granted, the agent orients itself against your project and is ready to use.

Adding the connector manually in Claude (by URL)

If AudiencePlayer MCP is not yet visible in your Claude directory, or your organization manages connectors itself, add it by URL. The server address is https://api.audienceplayer.com/mcp.

• Claude on the web, desktop or mobile (Free, Pro and Max plans): go to Customize > Connectors, click Add custom connector, enter the server address and click Add, then click Connect and sign in. If you are asked how to authenticate, choose Sign in now, and for the OAuth client choose Register automatically.
• Claude Team and Enterprise plans: an organization Owner goes to Organization settings > Connectors, selects Add, then Custom, and enters the server address. Each member then goes to Customize > Connectors, finds the connector labelled Custom, and clicks Connect to sign in with their own account.
• Claude Code: run claude mcp add --transport http audienceplayer https://api.audienceplayer.com/mcp, then run /mcp inside Claude Code to sign in.

Connecting ChatGPT and other MCP-capable agents (by URL)

Until the ChatGPT listing is approved, and for any other AI agent or client that supports remote MCP servers, connect directly. Follow your AI product's instructions for adding a remote MCP server (or custom connector) by URL, and enter the server address https://api.audienceplayer.com/mcp. The technical details for integrators:

• Server address: https://api.audienceplayer.com/mcp
• Transport: MCP over HTTPS with JSON responses (no server-initiated event stream), protocol version 2025-06-18.
• Sign-in: OAuth 2.0 authorization code flow with PKCE (S256) and dynamic client registration. Discovery documents: https://api.audienceplayer.com/.well-known/oauth-protected-resource and https://api.audienceplayer.com/.well-known/oauth-authorization-server. Requested scopes: api-admin-access and api-mcp-access (shown to the user as the "API MCP Access" consent).
• Unauthenticated requests are answered with HTTP 401 and a pointer to the discovery documents, so compliant clients start the sign-in automatically.
• If your client cannot complete the sign-in, for example because its redirect address is not accepted, contact AudiencePlayer support through this support portal.

3. What your AI agent can do

The tools available through this connector are generated directly from our live API, so we deliberately describe this in general terms rather than as a fixed feature list — new capabilities appear automatically as our API grows, without this article going stale. In short: everything you can do in the AudiencePlayer CMS, and more — practically the same reach as full AudiencePlayer API access. See AudiencePlayer API access for the complete picture of what that access level covers.

A few examples of what this looks like in practice:

• Full content management: Pages, Articles, Categories, Menus, and more.
• Full monetization management: Subscriptions, Products, Vouchers, and more.
• Full statistics/telemetry access: build your own reports and/or drill into usage, viewing, and revenue data.

A small set of sensitive operations are deliberately excluded regardless of what you ask for: login/token issuance, payment-provider connection and order refunds, and destructive user actions (deleting a user, forcing a password reset, project-privilege changes).

Example prompts

• "How many new subscriptions were purchased last month compared with the month before, split by Subscription, and how many UserSubscriptions are active today?"
• "Create a series called 'The Long Road' with 2 seasons of 6 episodes each, add a video Asset to every episode, and upload the episode files from my folder."
• "List the Articles in the category Documentaries that are not published yet."
• "Create a Category called Drama and move these three Articles into it."
• "Show me the 10 most watched Articles last month."
• "Create a Subscription called Premium Monthly and show me which Products it includes."
• "Which live-video Assets are active right now?"
• "Create a new Page for our summer campaign and add it to the main Menu."

4. Full admin environment access: Always read before enabling

5. Uploading files through an agent

Whether file uploads work directly depends on the AI tool itself, not on AudiencePlayer: an agent running in a "coding" context (with direct file/network access) can typically upload a file straight through the connection. An agent running in a pure "chat" context (e.g. Claude.ai or ChatGPT’s web/app chat) generally cannot send raw file bytes reliably today — in that case, the agent will instead offer you a separate, short-lived upload link to complete the upload yourself. This is a current limitation of the broader MCP ecosystem across chat-based AI products, not something specific to AudiencePlayer’s implementation.

The same applies to large video/audio Asset uploads: if your agent supports direct upload, large media files can go straight through it; if not, use the AudiencePlayer CMS directly for those uploads.

For images, use PNG (when transparency is needed) or JPG; SVG is not recommended.

6. Staying in control

• Review before you approve. Your AI tool shows what it is about to do. Read approval requests carefully, especially for deletions, starting a live stream (this typically incurs streaming costs), and anything that changes Subscriptions or entitlements.
• Destructive actions are labelled. Tools that change or delete data are marked as such, so AI tools that support this ask for your approval before running them.
• Block tools you do not need. Most AI tools let you disable individual tools. In Claude, go to Customize > Connectors, select the AudiencePlayer connector, and set individual tools to Blocked.
• Switch the connector off when you do not need it. In Claude, use the + menu in a chat, choose Connectors, and turn AudiencePlayer off for that conversation.
• Use "Always allow" sparingly, and only for tools you fully trust.
• Sign-in is personal. Everyone connects with their own administrator account.

7. Usage limits

• Request rate: up to 120 requests per minute. Requests beyond that are answered with a "too many attempts" error; wait a minute and ask again.
• Size of results: AI tools cap how much a single result can contain (Claude on the web and desktop around 150,000 characters; Claude Code 25,000 tokens by default). List results are paginated, so for large projects ask for a filtered or smaller selection (for example "the 20 most recent Articles") instead of everything at once.
• Files: see section 5 for uploads and the recommended route for large media.

8. Privacy and support

• What the agent can see: the data your administrator account can see, including your users’ personal data. Only connect AI tools you trust.
• Not available through the connector: login and token issuance, password resets, payment-provider connections, order and payment creation, refunds, and deleting users.
• Support: if something does not work, contact AudiencePlayer support through this support portal and include the time of the problem and the name of the tool the agent was using.