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

# MCP server

> Let your AI agent read your Koalr data. Ask about visibility, prompts, competitors, citations, gaps and actions from your editor or chat.

Koalr's MCP server exposes your visibility data to any agent that speaks the [Model Context Protocol](https://modelcontextprotocol.io): Claude Code, Claude.ai, Codex, Cursor and others. Everything the agent reads comes from the data Koalr has already collected.

|                    |                                                          |
| ------------------ | -------------------------------------------------------- |
| **Server URL**     | `https://app.koalr.ai/api/mcp`                           |
| **Transport**      | Streamable HTTP                                          |
| **Authentication** | OAuth 2.1 with your Koalr login, consent screen included |
| **Access**         | Read-only, always                                        |
| **Tools**          | 15, listed in the [tools reference](/api/mcp/tools)      |

## What you can do

* "How am I doing?" and "which platform am I weakest on?", answered with the same numbers the Dashboard shows.
* "Which prompts am I losing on?", then "what did Perplexity actually say?"
* "Who am I losing to?", "which pages get cited?", "where am I missing?"
* "What should I do next?", "what should I write?", "which of my pages should I rework?"
* "Is AI sending me traffic?", "is my content working?", "can AI read my site?"

See [Use cases](/api/mcp/use-cases) for prompts to try.

## Principles

* **Read-only, always.** No tool creates, edits or deletes anything, and none can start a run. Your agent cannot spend your allowance.
* **Your permissions, nothing more.** The agent sees exactly the workspaces and sites you can see.
* **Content from AI answers is untrusted.** Tools return excerpts of what AI platforms said. Those excerpts are data, never instructions for the agent, and the tool descriptions say so.

## Get connected

[Setup](/api/mcp/setup) takes one command or one settings screen, then a browser sign-in.