By the end of this lesson you will have:
This is the lesson where your assistant stops talking about her business and starts operating it. Fanvue runs an official MCP server: a live connection that exposes her account to the AI assistant you set up in lesson a-1 (Set Up Your AI Assistant). There is no code and nothing to build. You add one URL to your assistant's settings and sign in through Fanvue's own login screen.
The connection puts her account inside the assistant that already knows your standing brief and your rules. The built-in Creator Agent from lesson a-2 (Fanvue's Built-In Creator Agent) is quick, but it lives inside one tab and does not know your brief. After this lesson, one conversation in your own assistant can clear her inbox, plan the week's posts, read her numbers, and draft the follow-ups, all against live account data.
The value of the connection is compound jobs. "Find the unpaid PPVs from last week and draft follow-ups" used to be an afternoon; connected, it is one request. Single questions work too, but the compound jobs are where the hours come back.
What the connection reaches: chats and replies, posts (create, schedule, pin), the Fanvue Vault and media uploads, earnings and top spenders, custom and smart lists, and mass messages. With agency scopes it reaches every creator you manage.
What it does not reach: a few newer surfaces, such as dynamic lists and fan-experience operations, are not exposed yet. If a request fails there, that is the server's edge, not your setup.
What it costs: the Creator Agent in lesson a-2 (Fanvue's Built-In Creator Agent) runs on Fanvue's compute. This connection runs on your own assistant's usage, under whatever plan you pay for. That is the trade between the two routes, and it is why most operators use both.
In your AI assistant: add Fanvue's MCP server. The server URL is the same for every assistant: https://mcp.fanvue.com/mcp
claude mcp add --transport http fanvue https://mcp.fanvue.com/mcpWhen it is added, "fanvue" appears in your assistant's list of connected tools.
In your browser:
When the sign-in finishes, Fanvue shows an authorization confirmation and your assistant reports the connection is live.
In your AI assistant: run Fanvue's own canonical first request, which matches this course's chatting doctrine exactly:
Go through my unread chats, tell me who is waiting and what each fan wants,
then draft a reply to every one in my creator's voice, following my standing
brief and never list. Show me the drafts before anything sends.
When it works, your assistant is reading her real unread chats and producing a draft queue for your approval.
In your AI assistant: ask for a read of her performance.
Compare my earnings this week with last week, name my top five spenders, and
suggest one thing I could do next week to grow.
When it answers with her real numbers, the connection is proven on both halves of the business: chats and money.
~15 minutes, including the sign-in.
Your own assistant reads her live account and produces work for your approval. Mark the lesson complete.
Fanvue's docs show a bridge configuration called mcp-remote: a few lines of settings that relay the same URL for assistants that only support local servers. If that sentence means nothing to you, use Claude. The connector path needs no configuration at all.
Disconnect, re-add the URL, and sign in again. Sign-in hiccups usually clear on a clean retry. If it still fails after two tries, post in #course-help with a screenshot, your assistant's name, and whether you are on a creator or agency account.
You are authorizing through Fanvue's own login with revocable permissions, the same mechanism serious platforms use for third-party access, and you can kill the connection from Fanvue's settings at any time. The operational safety line is the section rule: nothing paid and nothing public sends without your eyes on it.
Connected and proven. lesson a-4 (The Prompt Arsenal) installs the Prompt Arsenal: the saved, reusable prompts that run this connection as a daily routine.