> ## Documentation Index
> Fetch the complete documentation index at: https://api.fanvue.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Build with AI

> Two ways to put AI to work on Fanvue: operate your account in plain language, or have a coding agent build your integration.

There are two ways to bring AI to Fanvue, and they solve different problems. Pick the one that matches your goal.

<CardGroup cols={2}>
  <Card title="Operate with AI (no-code)" icon="wand-magic-sparkles" href="/docs/mcp-server">
    Connect the hosted Fanvue copilot and run your account in plain language. Reply to fans, pull insights, manage and schedule posts, all from an assistant like Claude. Nothing to build: you connect to `https://mcp.fanvue.com/mcp`, authorize with Fanvue, and start asking. Ideal for agency operators who want results without code.
  </Card>

  <Card title="Build with AI (coding agents)" icon="code" href="/docs/build-with-ai/quickstart">
    Point your coding agent (Cursor, Claude Code) at the Fanvue docs and API. It searches the documentation and calls real endpoints while it writes your integration, so you ship faster with fewer dead ends. For developers building their own app.
  </Card>
</CardGroup>

## Which one is for you

<Note>
  One distinction makes the choice obvious. Connecting the hosted copilot needs **no app of your own**: you just authorize and go. The moment you want your **own** integration, you're [building an app](/docs/app-store/introduction), because that is how Fanvue issues your API credentials.
</Note>

## Give your coding agent the docs and API

Building well with AI comes down to one thing: stop your agent guessing. Feed it Fanvue, and the more you connect, the better it works. Use any or all of these.

<CardGroup cols={2}>
  <Card title="llms.txt" icon="file-lines" href="/docs/build-with-ai/docs-for-your-llm">
    Point your AI at the docs as plain text. `llms.txt` is the index, `llms-full.txt` is the entire docs in one file. Drop the URL into any agent that takes context. For many builds, this alone is enough.
  </Card>

  {/* HIDDEN — future feature (Docs MCP). Remove this comment wrapper to restore.
    <Card title="Docs MCP" icon="magnifying-glass" href="/docs/build-with-ai/quickstart">
      Let a coding agent search the live docs as it writes, so it follows the current OAuth flow, scopes, and conventions instead of a stale copy.
    </Card>
    */}

  <Card title="API MCP" icon="plug" href="/docs/mcp-server">
    Let the agent call and test real endpoints as it builds, authorized through OAuth, with no keys to paste.
  </Card>
</CardGroup>

<Tip>
  Fastest path: point your agent at `llms-full.txt`, start from the [App Starter](https://github.com/fanvue/fanvue-app-starter), and paste the [golden prompt](/docs/build-with-ai/quickstart). Add the API MCP server when you want real API calls. Running both modes is common: operate day to day with the copilot, and build the rest with a coding agent.
</Tip>