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

# Connect via MCP

> Use Larm from Claude Code, Cursor, or any MCP-compatible AI tool

Larm has a built-in [MCP](https://modelcontextprotocol.io) server that lets AI tools manage your monitors, check uptime, create disruptions, and set up alerts — all through natural language.

Available on all plans, including Free.

## Claude Code

Run this command in your terminal:

```bash theme={null}
claude mcp add larm --transport http https://app.larm.dev/mcp
```

Or add it to your project's `.mcp.json`:

```json theme={null}
{
  "mcpServers": {
    "larm": {
      "type": "http",
      "url": "https://app.larm.dev/mcp"
    }
  }
}
```

## Claude Desktop

Open **Settings > Developer > Edit Config** and add:

```json theme={null}
{
  "mcpServers": {
    "larm": {
      "type": "http",
      "url": "https://app.larm.dev/mcp"
    }
  }
}
```

## Cursor

Open **Settings > MCP Servers** and add a new server:

* **Name:** `larm`
* **Type:** `http`
* **URL:** `https://app.larm.dev/mcp`

Or add to your project's `.cursor/mcp.json`:

```json theme={null}
{
  "mcpServers": {
    "larm": {
      "type": "http",
      "url": "https://app.larm.dev/mcp"
    }
  }
}
```

## VS Code

Add to your project's `.vscode/mcp.json`:

```json theme={null}
{
  "servers": {
    "larm": {
      "type": "http",
      "url": "https://app.larm.dev/mcp"
    }
  }
}
```

<Note>
  VS Code uses `servers` as the root key, not `mcpServers`. MCP tools work in Copilot's Agent mode.
</Note>

## Authentication

The first time you use a Larm tool, your MCP client will open a browser window:

1. Log in with your Larm account (email, GitHub, or Google)
2. Approve the authorization request — "Full access to your Larm account"
3. Return to your editor — the connection is ready

Tokens refresh automatically. You won't need to authorize again unless you revoke access.

## Available tools

### Read

| Tool                     | Description                                            |
| ------------------------ | ------------------------------------------------------ |
| `list_monitors`          | List all monitors with their current state             |
| `get_monitor`            | Full details: config, state, uptime %, p95, error rate |
| `get_uptime_summary`     | Fleet-wide uptime, disruption count, and MTTR          |
| `list_alert_channels`    | All configured alert channels                          |
| `list_state_transitions` | Monitor state changes in the last 24 hours             |

### Write

| Tool                   | Description                                              |
| ---------------------- | -------------------------------------------------------- |
| `create_monitor`       | Create a new HTTP monitor                                |
| `update_monitor`       | Change URL, interval, timeout, or enable/disable         |
| `create_alert_channel` | Create a webhook, email, Discord, or other channel       |
| `test_alert_channel`   | Send a test notification                                 |
| `list_disruptions`     | List status page disruptions                             |
| `create_disruption`    | Create a disruption, optionally publish to a status page |
| `update_disruption`    | Post updates, advance status, or resolve                 |

## Examples

Ask your AI tool:

* "What's the uptime of my monitors this week?"
* "Create a monitor for [https://api.example.com/health](https://api.example.com/health)"
* "Set up a Discord webhook for alerts"
* "Create a disruption — the database is down — and publish it to the production status page"
* "Resolve the database disruption"
* "Pause the staging monitor"

## Rate limits

60 requests per minute per token. If you hit the limit, wait 60 seconds.