Setter Books
Sign inStart a book →

For people who would rather ask

Make a book from your AI assistant.

Connect Setter Books to Claude, ChatGPT or another assistant, and make a book by talking it through: who is in it, what happens, how it should look. It starts the free proof, reads you the story, shows you the pages and changes what you ask. When it looks right, it hands you the link to pay.

It is the same account as the website: books you start from your assistant are on your shelf here, and the other way round. You need an account to connect; making one is free.

Connect it

Each assistant does it a little differently. The first time, it opens a Setter Books page where you sign in and allow it. From then on it acts as you: it sees your books and characters, and nobody else's.

  • Claude

    claude.ai, the desktop app, Cowork

    Open Customize → Connectors, press +, then Add custom connector. Call it Setter Books, paste the address and add it. Press Connect and sign in to Setter Books when it asks. In a chat, switch it on under + → Connectors. Every plan can add a custom connector (the Free plan allows one). On Team and Enterprise an owner adds it once under Organization settings → Connectors (Add → Custom → Web), and then each person connects it for themselves.

    The address
    https://setterbooks.com/mcp

  • Claude Code

    in the terminal

    Add the server, then run /mcp inside Claude Code, pick setter-books and sign in.

    In your terminal
    claude mcp add --transport http setter-books https://setterbooks.com/mcp

  • ChatGPT

    on the web, in developer mode

    For now ChatGPT only connects to a server like ours in developer mode, which is on the Plus, Pro, Business, Enterprise and Education plans, on the web. Turn it on in Settings → Security and login → Developer mode. Then open Plugins, press + to make a developer-mode app, paste the address, choose OAuth, and sign in to Setter Books when it asks.

    The address
    https://setterbooks.com/mcp

  • Codex

    the CLI, the IDE extension, the app

    The second line opens the sign-in. The CLI, the IDE extension and the Codex app share these settings, so once is enough.

    In your terminal
    codex mcp add setter-books --url https://setterbooks.com/mcp
codex mcp login setter-books

  • Cursor

    the editor

    Put this in ~/.cursor/mcp.json (or .cursor/mcp.json for one project), then sign in to Setter Books when Cursor asks. Or add it to Cursor in one click.

    mcp.json
    {
  "mcpServers": {
    "setter-books": {
      "url": "https://setterbooks.com/mcp"
    }
  }
}

  • VS Code

    with GitHub Copilot

    Put this in .vscode/mcp.json, or in your own settings with the MCP: Open User Configuration command. Start the server from the file and sign in when asked; Copilot's agent mode can then use it.

    mcp.json
    {
  "servers": {
    "setter-books": {
      "type": "http",
      "url": "https://setterbooks.com/mcp"
    }
  }
}

Anything else

An assistant that connects to MCP servers over HTTP and signs in with OAuth can use the same address. Gemini CLI, for one:

In your terminal
gemini mcp add --transport http setter-books https://setterbooks.com/mcp

Then just say what you want

It asks what it needs to know: who is in the book, what happens, the art style, how long. A proof takes a few minutes to write and paint; ask how it's getting on.

  • “Make a picture book for Maya. She's five, she wears yellow boots everywhere, and the squirrels next door have taken one of them.”
  • “Read me the story. On page 3, have Grandpa say it instead.”
  • “Show me the cover and the first two pages.”
  • “What would the hardcover cost, sent to my sister in Toronto?”

What it can do, and what it can't

It can

  • · Add the people in your family (and the dog) as characters, from what you tell it, and correct how they are drawn.
  • · Start a free proof: it writes the story, draws the cast, and paints the cover and the first two spreads.
  • · Read you the story, show you the pictures (watermarked until you order), and change any words or move them to another corner. While a hardcover order can still be cancelled, it leaves the finished pages for you to open here, because opening them ends that chance.
  • · Repaint a page or the cover, or make one change to a picture, when you ask. Approve pages (an approved page is final), or pick between two versions.
  • · Tell you what a hardcover costs with shipping to an address, and make the link to pay.
  • · Tell you where an order is, and cancel a hardcover before it goes to the printer.

It can't

  • Pay. When you are ready it gives you a link to Stripe's checkout page, and you pay there yourself. It never sees or types card details. A hardcover is a hold on your card that we only charge once the printer takes the job; a PDF is charged straight away.
  • Send photos. It gives you a link to a page here where you add them from your phone or computer. Photos make the likeness; without them we draw someone from the description.
  • Get round the limits. It uses your account, so the same rules apply: three free proofs a day, four repaints a book before you order, and the same line on what we will and won't make.
  • Delete anything. Books and characters are deleted here, on the site.

Anything that uses something up (a free proof from today's allowance, a repaint, a checkout) is marked that way for the assistant, and it is told to check with you first. Most assistants also ask your permission before they use a tool that changes something; you can usually set them to ask every time.

What the assistant sees

What it reads through Setter Books (names and descriptions, your story, the pictures it looks at, your orders), and the name, email address and profile picture that signing in shares, go to the company that makes the assistant, and are handled under your agreement with them. We record what it does the way we record what you do on the site, never with your words or photos. To stop, remove Setter Books from your assistant's connectors. The details are on the privacy page.

In the browser

Some browsers are starting to let an AI agent use a website's own tools, through a new standard called WebMCP. When you are signed in, our pages offer theirs: an agent can fill in the new-book form, turn the pages of your book, fix a line or fill in a shipping address while you watch, and anything that spends something asks you on the page first. It is early. It works in Chrome while Google trials WebMCP (or with its testing flag switched on), few agents use it yet, and nothing changes for anyone else.