# Deployment

> How to roll out the Holistic AI EndLayer agent: one device in minutes, a fleet through Intune, Jamf or SCCM, supported platforms and network needs.

Rolling out Holistic AI EndLayer, the Guardian Agent Platform, is two steps: install the Layer 1 Agent, then connect it to your organisation. Only the second step changes as you scale from one device to thousands.

## Choose the rollout

| Rollout | Use it when | How a device joins |
|---|---|---|
| **One device** | A pilot, or your own laptop | A short claim code, approved in the console |
| **A fleet** | Jamf, Kandji, SCCM, configuration management, or Intune by hand | One enrolment secret, with nobody needed at each device |
| **From the console** | You use Intune or Jamf | EndLayer pushes the agent to the groups you pick. See [integrations](https://www.endlayer.ai/docs/connections). |
| **Containers and build machines** | Short-lived machines | The enrolment secret in a file |

## Supported platforms

| Platform | Supported versions | Package |
|---|---|---|
| **macOS** | macOS 12 and later | A universal installer, signed and notarised by Apple |
| **Windows** | Windows 10 and Windows Server 2019 and later | A per-machine installer with silent install |
| **Linux** | Debian, Ubuntu, RHEL, Fedora and SUSE | Packages for Intel and Arm, plus a plain archive |

Phones and tablets are listed through Microsoft Intune rather than an agent.

Installing needs administrator rights once. Running needs none: the agent runs as the person whose device it is. Installers come from your own organisation's console, and every download can be checked against a signed checksum manifest.

## One device, in about ten minutes

1. **Install:** Download the installer from your console and run it.
2. **Point it at your organisation:** Add your organisation's EndLayer address to the agent's settings. Until you do, nothing leaves the device.
3. **Connect it:** Run `endlayer enroll` as the person who uses the device. It shows a short code, and somebody signed in to your organisation types it into the console.
4. **See the first report:** The device appears in the console after its first scan. If anything looks wrong, run `endlayer doctor` first.

## A fleet, with nobody at any device

**Connections, Fleet install** creates one enrolment secret, with a device limit and an expiry you choose, and generates install scripts for Windows, macOS and Linux with your organisation's details already filled in. The secret is shown once. The page then shows enrolments per secret, so a rollout that stalls partway is visible at a glance.

Devices installed by hand can also show a small menu-bar or tray app: whether the device is connected, and the warnings and blocks its person has run into. It has nothing that changes the rollout.

> **On macOS, send the profiles with the package:** Push a login-items profile and a folder-access profile through your device management tool alongside the installer. The first keeps macOS from offering each person a switch to turn the agent off; the second lets the agent read the folders where projects live. The Intune and Jamf routes push both for you. For any other tool, they are downloads beside the install scripts.

## What it asks of a network

| Need | Detail |
|---|---|
| **Destination** | One place, your organisation's EndLayer service, on port 443 outbound |
| **Inbound** | None. Nothing listens on the device. |
| **Volume** | A few tens of kilobytes per report, every 15 minutes |
| **TLS inspection** | Supported, with your root certificate added. Certificate checks stay on. |

## The browser extension

The browser extension checks prompts, uploads and downloads on AI websites. Your device management tool installs it in Chrome, Edge, Brave, Vivaldi or Chromium, and it hands each check to the agent on the same device. Firefox support is planned.

## After the first devices report

Until **policy baselines** are set, most checks report and stay ungraded on purpose. That is the honest answer to "does this match policy" when nobody has said what the policy is. Set these first:

- **Corporate domains:** so accounts can be told apart as corporate or personal.
- **Agent modes you permit:** for example, whether coding agents may skip permission prompts.
- **Domains browsers must block:** the AI websites your policy rules out.
- **What you pay per tool:** so seats can be set against use.

Then turn on the prompt scanner at Monitor, and move rules to Warn, Redact or Block once the events show what they would catch. See [what it stops](https://www.endlayer.ai/docs/enforcement).

  - [Integrations](https://www.endlayer.ai/docs/connections): Intune, Jamf, cloud accounts and alert destinations.
  - [Trust model](https://www.endlayer.ai/docs/trust): How devices prove who they are.

Source: https://www.endlayer.ai/docs/deployment · Updated 2026-11-20
