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

# Cited MCP

> Connect Cited to your AI client (Claude, Cursor, ChatGPT) and query your AI-visibility data and Cited's execution layer with an MCP server.

The **Cited MCP server** exposes your brand's AI-visibility data — and, over time, Cited's execution layer — as tools your own AI client can call. Instead of switching to the Cited dashboard, you ask your assistant (Claude Desktop, Cursor, ChatGPT, Claude Code) a question and it calls Cited directly.

Cited MCP is built on the open [Model Context Protocol](https://modelcontextprotocol.io) — the standard for connecting AI applications to external systems.

## Who it's for

* **Agencies** managing many brands — one API key can be granted **multiple brands**, so an assistant can loop over your whole portfolio ("for each brand, pull this week's gaps and summarize").
* **In-house power marketers** who live in an AI tool and want programmatic, conversational access to their AI-visibility data alongside their other work.

<Note>
  Cited MCP is an **agency / power-user surface**. If you just want to look at your numbers, the [dashboard](https://dashboard.getcited.in) is faster and richer. The MCP shines for automation and for composing Cited data with your other tools inside an agent.
</Note>

## What it is (and isn't)

Cited MCP is the **external, agent-facing face of the Cited platform**. It is a thin projection — every tool returns the same data the dashboard shows.

* **Available now:** resolver tools (`whoami`, `list_brands`, `list_dimensions`) and read tools (`get_ai_shelf_products`, `get_cited_index`). See [Capabilities](/mcp/capabilities).
* **Coming:** conformed **metric tools** (`cube_slice` / `cube_drill`) that return the exact numbers from Cited's metrics cube, and **action tools** (generate a content brief, find source targets, create a task) that run Cited's execution layer.

## How it works

There are two ways to connect, depending on your client:

* **Sign-in (OAuth)** — on **Claude Desktop**, paste one URL (`https://api.getcited.in/mcp-oauth`) and **sign in with your Cited account**. No key to manage; your brands are resolved from your verified email.
* **API key** — on every other client (Claude Code, Cursor, VS Code, Gemini CLI, ChatGPT), authenticate with a **Cited API key** (`Authorization: Bearer cited_sk_…`) against **`https://api.getcited.in/mcp`**.

Either way, your assistant then calls Cited tools, and **each call is scoped to the brands your identity or key is granted**.

Start with [Connect](/mcp/connect) to wire up your client, then see [Capabilities](/mcp/capabilities) and [Recipes](/mcp/recipes).

<Warning>
  Connecting the MCP sends the Cited data your assistant requests to **whatever AI client you connect** (e.g. Claude Desktop, ChatGPT). It exposes only data you already have access to in Cited — but that data leaves Cited for your chosen client. Treat your API key like a password.
</Warning>