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

# Connect Endpoint

> The three steps that put a machine into Fleet — install the package, issue an endpoint key, enroll — and which package to use for each operating system.

## What this is for

**Connect endpoint** is the button at the top right of **Endpoint AI Risk › Fleet** (it also opens
from the empty fleet, from Inventory and from the Skills Marketplace). It walks through the three
things a machine needs before it reports: the agent installed, an endpoint key, and one enroll
command. Fleet administrators use it; the same steps work for one laptop by hand or for a thousand
through an MDM.

## The three steps

<Steps>
  <Step title="Install gw-cli">
    The dialog offers a download for every platform, with the machine you are browsing from preselected. Pick the package, not the archive, for anything that is not an evaluation:

    | Operating system | Package | How it is deployed |
    | - | - | - |
    | macOS 13 or later, Apple Silicon and Intel | `guardway-gw-cli-macos-universal.pkg` | Jamf, Kandji, Intune, Mosyle, Workspace ONE, Addigy, or `installer -pkg` |
    | Windows 10 and 11, x86\_64 | `guardway-gw-cli-windows-x86_64.msi` | Intune, SCCM, Group Policy, or `msiexec /i` |
    | Ubuntu 22.04+, Debian 12+ | `guardway-agent_<version>-1_amd64.deb`, `_arm64.deb` | `apt install`, or Ansible, Puppet, Chef, Salt |
    | RHEL, Rocky and Alma 9, Fedora | `guardway-agent-<version>-1.x86_64.rpm`, `.aarch64.rpm` | `dnf install`, or your config manager |

    Installing the package is the whole install. It starts the agent for every signed-in user right away and at every sign-in after that, in the **Observe** posture. See [macOS](/endpoint-ai-risk/connect/macos), [Windows](/endpoint-ai-risk/connect/windows) and [Linux](/endpoint-ai-risk/connect/linux).
  </Step>

  <Step title="Issue an endpoint key">
    Type a **Machine label** (for example `dana-mbp`) and click **Generate key**. The key begins with `sk-ep-`, enrolls one machine, and is shown once: copy it before you close the dialog. Issue one key per machine so you can tell them apart in Fleet.
  </Step>

  <Step title="Enroll the machine">
    The dialog fills your key into the command. Run it on the machine as the signed-in user:

    ```sh theme={null}
    gw-cli enroll --key sk-ep-… --url https://api.us.guardway.ai --verify
    ```

    `--verify` sends a first report straight away. The machine appears in Fleet within a minute. Details, including what to run on a managed install and how to script it, are on [Activate](/endpoint-ai-risk/connect/activate).
  </Step>
</Steps>

## Which install do I want?

| | Package (MDM or by hand) | Archive (`.tar.gz`, `.zip`) |
| - | - | - |
| Who it is for | Every customer machine | Evaluation on your own laptop |
| Where it installs | Machine-wide, for every user | The current user's home folder |
| Starts by itself | Yes, at every sign-in | Yes, for that user only |
| Can enforce policy | Yes: Live guard Block, AI Control, skill custody, Enforce posture | Records only |
| Updates | Through the console's update policy, or the next package | `gw-cli upgrade` |

The archives carry the same binary and a per-user `install.sh` or `install.ps1`. They are fine for
trying the agent, and the [Agent overview](/endpoint-ai-risk/agent/overview#evaluation-install) covers
them. Use the packages for anything you will manage.

## Downloads and checksums

Every release is published at `https://api.guardway.ai/v1/downloads/gw-cli/<version>/<file>`, and the
newest at `.../gw-cli/latest/<file>`. A `SHA256SUMS` file sits next to the packages in each folder;
the Linux packages also ship a `.sha256` each. Verify before you push a package to a fleet:

<CodeGroup>
  ```sh macOS / Linux theme={null}
  curl -LO https://api.guardway.ai/v1/downloads/gw-cli/latest/SHA256SUMS
  curl -LO https://api.guardway.ai/v1/downloads/gw-cli/latest/guardway-gw-cli-macos-universal.pkg
  shasum -a 256 -c --ignore-missing SHA256SUMS
  ```

  ```powershell Windows theme={null}
  Invoke-WebRequest https://api.guardway.ai/v1/downloads/gw-cli/latest/SHA256SUMS -OutFile SHA256SUMS
  Invoke-WebRequest https://api.guardway.ai/v1/downloads/gw-cli/latest/guardway-gw-cli-windows-x86_64.msi -OutFile guardway-gw-cli-windows-x86_64.msi
  (Get-FileHash .\guardway-gw-cli-windows-x86_64.msi).Hash.ToLower()   # compare with the SHA256SUMS line
  ```
</CodeGroup>

<Note>
  The download buttons in the console appear once downloads are switched on for your organization.
  Until then the dialog shows the archive command, and the URLs above still work.
</Note>

## Before a fleet rollout

* Allow the hosts on [Network Requirements](/getting-started/network-requirements#endpoint-agent) on each machine. The agent ignores proxy settings and trusts only public certificate authorities.
* On macOS, push a Full Disk Access profile with the package; without it the agent cannot read `~/Desktop`, `~/Documents` and `~/Downloads`. See [macOS](/endpoint-ai-risk/connect/macos#full-disk-access).
* Decide the install mode. Every package installs **Observe** by default, which never changes how a developer's tools behave. The **managed** install adds the root-level service that Enforce, Live guard Block and skill custody need. [Activate](/endpoint-ai-risk/connect/activate#install-modes) explains how to choose it per OS.
* Create an [Endpoint AI Policy](/endpoint-ai-risk/policy) for the machines before they arrive, or they take the policy bound to **All endpoints**.

## Related

<CardGroup cols={2}>
  <Card title="Activate" icon="key" href="/endpoint-ai-risk/connect/activate">
    Enroll, verify the first report, and choose the install mode.
  </Card>

  <Card title="Agent updates" icon="arrows-rotate" href="/endpoint-ai-risk/connect/updates">
    How machines move to a new agent release.
  </Card>

  <Card title="Fleet" icon="laptop" href="/endpoint-ai-risk/fleet">
    Where an enrolled machine shows up.
  </Card>

  <Card title="Uninstall & Recovery" icon="trash-can" href="/endpoint-ai-risk/agent/uninstall">
    Removing the agent in the safe order.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.