---
title: MCP server
description: Give AI platforms direct access to Array HQ capabilities through the Model Context Protocol — query metrics, explore organizational data, and build AI-driven workflows.
product: platform
audience: operator
voice_pillar: coordinate
---

# MCP server

> Give AI platforms direct access to Array HQ capabilities through the Model Context Protocol.

## The operational problem

AI platforms can analyze data and automate workflows — but only if they can reach the systems where that data lives. Without a standard way to connect, every AI integration requires custom scripting, manual data exports, or one-off API wrappers. Each new AI tool means another integration project.

## How Array HQ solves it

The Array HQ MCP server exposes platform capabilities through the Model Context Protocol (MCP) — an open standard that AI platforms use to discover and invoke tools. Connect your AI platform once, and it can query metrics, explore organizational data, and interact with Array HQ directly.

The server controls what each user can see and do. Tools are composed based on your permissions, so users only access the capabilities they're authorized to use.

## How it works

### Connecting an AI platform

AI platforms that support MCP (such as Claude, OpenAI, and others) connect to the Array HQ MCP server at its public endpoint. The connection uses the standard MCP protocol — no custom SDK or client library is required.

### Tools

The MCP server exposes platform capabilities as individual tools that AI platforms discover and invoke automatically through the MCP protocol. No configuration changes are needed when new tools become available — the AI platform picks them up on the next connection.

Note that tools are currently provided primarily for the //Ops member role. More tools will be added over time to support additional domains and roles.

### Authentication & Authorization

The MCP server uses OIDC authentication through `auth.arraycorp.com` — the same identity provider used across Array HQ. This means the AI platform operates with the same access boundaries as the user — it cannot see or do anything the user couldn't do themselves through the standard interface.

Not every user sees every tool. The MCP server checks which capabilities are granted to the authenticated user and only exposes the tools they're authorized to use. This is based on the same permission policies that govern the rest of Array HQ.

### Setup Instructions

The Array HQ MCP server is available at:

```
https://mcp.arrayhq.com/mcp
```

#### Claude Code

Add the Array HQ MCP server to your Claude Code configuration using the `claude` CLI:

```bash
claude mcp add --transport http arrayhq https://mcp.arrayhq.com/mcp
```

The server uses OIDC authentication — Claude Code will prompt you to authenticate through the standard Array HQ login flow on first use.

#### Cursor

Add the Array HQ MCP server to your Cursor MCP configuration. Add the following to your `mcp.json` file (`.cursor/mcp.json` in your project directory, or `~/.cursor/mcp.json` for user-level configuration):

```json
{
  "mcpServers": {
    "arrayhq": {
      "url": "https://mcp.arrayhq.com/mcp"
    }
  }
}
```

Cursor will prompt you to authenticate through the standard Array HQ login flow when the server is first invoked.

### What you can build

With MCP access, AI platforms can:

- **Query workforce metrics** — ask natural-language questions about operational performance across sites and time periods
- **Analyze trends** — compare goal attainment, productivity, or staffing levels across organizations without manual data exports
- **Build reports** — generate operational summaries by pulling real-time data directly from Array HQ
- **Automate analysis** — set up recurring AI-driven reviews of metrics that surface signals and anomalies

## Related

- [AI Agents](../platform/ai-agents.md)
- [Accounts and Organizations](../platform/accounts-organizations.md)