Skip to main content

Getting Started with Swif's Agentic Security (for Enterprise)

If your organization already uses Swif for device management, you can turn on Agentic Security from your existing Swif workspace, with no separate sign-up.

Agentic Security gives your team visibility into the AI agents on your managed devices:

Feature

What it gives you

AI agent inventory

See which autonomous AI agents are installed across your fleet, such as OpenClaw, Hermes, and Grok Bot, and how they're set up

Risky setting detection

Find agents with settings that weaken their safeguards, such as an OpenClaw gateway exposed with no authentication

OpenClaw event logs

Review what OpenClaw did on each device, in near real time, for audits and investigations

AI coding agent monitoring

With the AI Agent Activity Monitoring Policy, review Claude Code, Codex, and Cursor sessions

Sensitive data redaction

API keys, passwords, and credentials are detected and hidden before they appear in your logs

CLI privilege protection

OpenClaw and other monitored CLI agents can't run with sudo or root

What you'll need

  • An active Swif enterprise workspace, with admin access

  • Devices running macOS, Windows, or Linux

Supported platforms

Platform

Versions

Installer package

Standalone installer

Architectures

macOS

macOS 12 or later

.pkg (universal), signed and notarized by Apple

installer-arm64.zip, installer-x64.zip

Apple silicon and Intel

Windows

Windows 10 or later

.msi, signed by Swifteam Inc.

installer-x64.exe, installer-arm64.exe

x64 and ARM64

Linux

Ubuntu 20.04+, Debian 11+, Fedora, RHEL / Rocky Linux / AlmaLinux 8+, openSUSE, with systemd

.deb (Debian, Ubuntu) or .rpm (Fedora, RHEL, Rocky Linux, AlmaLinux, openSUSE)

installer-x64, installer-arm64

x64 and arm64

Before you install

How installation works. Each installer installs agenticsecurity-install, a registration tool. Registering, with your Team ID and the device owner's details, downloads the Agentic Security agent, installs its background service, and enrolls the device. Without those details, the installer only installs the registration tool.

Don't run the agent-* files. Files named agent-x64, agent-arm64, or agent-x64.exe are the agent itself, not installers. Running one starts the agent in the window, which looks like it's stuck, and doesn't enroll the device. Always use the installer packages or standalone installers.

Use straight quotes. Type values in straight quotes ("), not curly quotes (“ ”). Curly quotes, which some apps add when you copy text, make enrollment fail.

Use the same region for the installer and the Team ID. Download the installer from the same Swif region, US or EU, as your workspace. To move a device to another region, remove the agent first.


Installing Agentic Security

There are three ways to install Agentic Security across your organization:

Method

Best for

Bulk deployment with Swif MDM (recommended)

Devices already managed by Swif

Silent deployment with another tool

Devices managed by another MDM, Microsoft Intune, or a Linux configuration tool

Manual install

Individual or BYOD devices

With bulk or silent deployment, each device installs the agent and enrolls automatically into your organization. Nothing needs to be done on the device itself.

Step 1: Open Agentic Security from your dashboard

Log in to your Swif workspace (app.swif.ai or app.eu.swif.ai). On your Home Dashboard, you'll see an Install Agentic Security card. You can also open it from the Setup Guide.

Bulk deployment with Swif MDM (recommended)

For devices managed by Swif, deploy Agentic Security to your whole fleet from the Agentic Security page.

  • Monitoring gap detection: an Agentic Security Monitoring Gap Detected alert appears at the top of the Agentic Security page when managed devices are missing the agent.

  • Configure policies: click Configure Policies in the alert to open Agentic Security Deployment Policies.

  • Platform-specific policies: configure each policy, or add devices to it:

  • Review missing devices: click Review Devices in the window to see every managed device that doesn't have the agent yet.

  • Auto-hide: the alert hides once every device in your organization has the agent.

Silent deployment with another tool

If your devices are managed by another MDM, Microsoft Intune, or a Linux configuration tool, deploy the installer package silently. When you supply your Team ID and the device owner's details, the device enrolls automatically during the install.

You'll need:

  • your organization's Team ID: Settings > Team

  • the device owner's first name, last name, and email

  • the installer package for each platform: Agentic Security > Devices > Setup guide

All four values are required. If any of them is missing, the installer stops with an error.

Supply each device owner's details

Each device needs its own owner's name and email. There's no single "fleet owner" value you can use for every device.

Deployment

How to supply each owner's details

Mac MDM, such as Jamf or Kandji

In the configuration profile, use your MDM's user variables for the assigned user's first name, last name, and email, instead of typed-in values. The MDM fills them in for each Mac. See your MDM's documentation for the variable names.

Microsoft Intune (Windows)

The line-of-business app's command line is the same for every device. For per-user values, deploy a script that looks up the signed-in user and runs the install.

Linux configuration tools

Generate each device's enroll.conf from a template that fills in the owner's details.

macOS: with an MDM

  1. Create a configuration profile with a Custom Settings payload:

    • Domain: com.swifteam.agentsecurity

    • Keys (all strings):

      • TeamId: Your Team ID

      • Name: The owner's first name

      • Surname: The owner's last name

      • Email: The owner's email

      • ForceEnroll (Boolean, optional): true to enroll a device again

  2. Deploy the profile to the Macs.

  3. Deploy the .pkg package. It installs the agent, enrolls the Mac, and starts the background service, without any prompts.

The package is signed by Developer ID Installer: Swifteam Inc (86XZ7Y7Z47) and notarized by Apple, so users don't see an "unidentified developer" warning.

macOS: without an MDM

Run these commands in Terminal on the Mac:

sudo mkdir -p "/Library/Application Support/AgenticSecurity"
sudo defaults write "/Library/Application Support/AgenticSecurity/enroll" TeamId "<TEAM_ID>"
sudo defaults write "/Library/Application Support/AgenticSecurity/enroll" Name "<FIRST_NAME>"
sudo defaults write "/Library/Application Support/AgenticSecurity/enroll" Surname "<LAST_NAME>"
sudo defaults write "/Library/Application Support/AgenticSecurity/enroll" Email "<EMAIL>"
sudo installer -pkg agenticsecurity-installer-<VERSION>-macos-universal.pkg -target /

Windows: silent install

From an administrator command prompt:

msiexec /i "agenticsecurity-installer-<VERSION>-windows-x64.msi" TEAMID="<TEAM_ID>" NAME="<FIRST_NAME>" SURNAME="<LAST_NAME>" EMAIL="<EMAIL>" /qn /l*v install.log

If enrollment fails, for example because the Team ID is wrong, the install fails with exit code 1603 and rolls back, so your deployment tool reports the failure.

The package is signed by Swifteam Inc.

Windows: Microsoft Intune

The steps below use the same values on every device. To give each device its own owner, deploy a script instead. See Supply each device owner's details.

  1. In Intune, add the .msi as a Line-of-business app.

  2. In the app's command-line arguments, enter:

    TEAMID="<TEAM_ID>" NAME="<FIRST_NAME>" SURNAME="<LAST_NAME>" EMAIL="<EMAIL>"

    Leave out /qn and /l*v. Intune adds them itself.

  3. Assign the app to your device groups.

Linux: with a configuration file

sudo mkdir -p /etc/agentsecurity
printf 'TEAM_ID=<TEAM_ID>\nNAME=<FIRST_NAME>\nSURNAME=<LAST_NAME>\nEMAIL=<EMAIL>\n' | sudo tee /etc/agentsecurity/enroll.conf
sudo apt install ./agenticsecurity-installer-<VERSION>-linux-x64.deb

On Fedora, RHEL, Rocky Linux, or AlmaLinux, use sudo dnf install ./agenticsecurity-installer-<VERSION>-linux-x64.rpm instead. On openSUSE, use sudo zypper install --allow-unsigned-rpm ./agenticsecurity-installer-<VERSION>-linux-x64.rpm. Use apt, dnf, or zypper, not dpkg -i or rpm -i, so required packages are installed too.

Linux: with environment variables

sudo env AGENTSECURITY_TEAM_ID=<TEAM_ID> AGENTSECURITY_NAME=<FIRST_NAME> AGENTSECURITY_SURNAME=<LAST_NAME> AGENTSECURITY_EMAIL=<EMAIL> dnf install ./agenticsecurity-installer-<VERSION>-linux-x64.rpm

The .deb and .rpm packages come with detached GPG signatures (.asc), and each release includes a SHA256SUMS file, so you can verify them before you deploy.

Devices that are already enrolled

Running the installer again on an enrolled device skips enrollment, and doesn't create a duplicate device. To enroll a device again, for example to move it to another team:

Platform

Add this, then reinstall

macOS

ForceEnroll (Boolean, true) in the profile or settings file

Windows

FORCEENROLL=1

Linux

FORCE_ENROLL=1 in enroll.conf

Send setup guides to device owners

To send individual instructions instead, use Review Devices on the Agentic Security page to email setup guides for each OS directly to device owners.

Step 2: Install the agent manually

(For single devices, testing, or BYOD)

  1. Click Setup Agentic Security from the dashboard card, or the Setup Guide.

  2. Choose the operating system and, for Windows and Linux, the processor type. For Linux, also choose the distribution.

  3. Choose the employee the device belongs to. This fills in their name and email in the install command, and turns on Download. Until you choose one, Download stays off.

  4. Download the installer, and run the command shown, on the device. Or send the employee a setup guide by email, with the command filled in.

The steps for each platform are below. Replace the values in <…> with your Team ID and the employee's details. All four are required.

macOS

With the .pkg package (recommended)

  1. Open the .pkg file, and follow the steps. Or, in Terminal:

    sudo installer -pkg agenticsecurity-installer-<VERSION>-macos-universal.pkg -target /
  2. Register the Mac. Use the full path to the registration tool:

    sudo /usr/local/bin/agenticsecurity-install -teamId="<TEAM_ID>" -name="<FIRST_NAME>" -surname="<LAST_NAME>" -email="<EMAIL>"

With the standalone .zip

Download the .zip for your Mac's processor, open Terminal in the folder that contains it, and run:

unzip installer-arm64.zip
chmod +x agenticsecurity/AgenticSecurityInstaller_arm64
sudo agenticsecurity/AgenticSecurityInstaller_arm64 -teamId="<TEAM_ID>" -name="<FIRST_NAME>" -surname="<LAST_NAME>" -email="<EMAIL>"

On an Intel Mac, use installer-x64.zip and AgenticSecurityInstaller_x64. For the EU region, the installer is named AgenticSecurityInstaller_eu_arm64 or AgenticSecurityInstaller_eu_x64.

Windows

Open PowerShell or Command Prompt with Run as administrator, in the folder that contains the installer.

With the .msi package, in one step (recommended)

msiexec /i "agenticsecurity-installer-<VERSION>-windows-x64.msi" TEAMID="<TEAM_ID>" NAME="<FIRST_NAME>" SURNAME="<LAST_NAME>" EMAIL="<EMAIL>"

This installs the registration tool, and registers the device. Use the uppercase properties shown, not -teamId and the other options used below. If any value is missing, the install rolls back.

With the .msi package, in two steps

  1. Double-click the .msi file, and follow the steps. It installs agenticsecurity-install.exe, and adds it to the system path.

  2. Open a new administrator terminal, and run:

    agenticsecurity-install.exe -teamId="<TEAM_ID>" -name="<FIRST_NAME>" -surname="<LAST_NAME>" -email="<EMAIL>"

With the standalone .exe

.\installer-x64.exe -teamId="<TEAM_ID>" -name="<FIRST_NAME>" -surname="<LAST_NAME>" -email="<EMAIL>"

On ARM64 devices, use installer-arm64.exe.

Linux

With the .deb or .rpm package (recommended)

  1. Install the package. Use apt, dnf, or zypper, not dpkg -i or rpm -i, so required packages, such as sudo, are installed too.

    • Debian, Ubuntu:

      sudo apt install ./agenticsecurity-installer-<VERSION>-linux-x64.deb
    • Fedora, RHEL, Rocky Linux, AlmaLinux:

      sudo dnf install ./agenticsecurity-installer-<VERSION>-linux-x64.rpm
    • openSUSE:

      sudo zypper install --allow-unsigned-rpm ./agenticsecurity-installer-<VERSION>-linux-x64.rpm

    On arm64 devices, replace x64 with arm64 in the package name.

  2. Register the device. Use the full path to the registration tool:

    sudo /usr/sbin/agenticsecurity-install -teamId="<TEAM_ID>" -name="<FIRST_NAME>" -surname="<LAST_NAME>" -email="<EMAIL>"

With the standalone installer

chmod +x installer-x64
sudo ./installer-x64 -teamId="<TEAM_ID>" -name="<FIRST_NAME>" -surname="<LAST_NAME>" -email="<EMAIL>"

On arm64 devices, use installer-arm64.

Check that a device is enrolled

OS

Run this

You should see

macOS

sudo launchctl print system/com.swifteam.agentsecurity-agent | grep state

state = running

Windows

Get-Service STAgentSecurity (in PowerShell)

Running

Linux

systemctl status agentsecurity-agent

active (running)

The device also appears in your Agentic Security device list. Once the Swif AI Agent Activity Monitoring Policy is on, sessions from Claude Code, Codex, and Cursor appear within about 15 minutes. See Step 3: Turn on your policies.

Logs

OS

Logs

macOS

/var/log/install.log, /var/log/agentsecurity.log

Windows

The MSI install.log, C:\Program Files\AgentSecurity\logs\log.txt

Linux

apt or dnf output, /var/log/agentsecurity.log, journalctl -u agentsecurity-agent

Messages to look for: Enrollment Successful, device already enrolled, no TeamId configured, and Installation failed: team … is not valid.

The installer seems stuck

You may have run an agent-* file, such as agent-x64, instead of an installer. It starts the agent in the window, and doesn't enroll the device. Press Ctrl+C, then use an installer package or standalone installer.

Enrollment fails, though the values look right

  • Check for curly quotes (“ ”). Retype them as straight quotes (").

  • Check that the installer and the Team ID are from the same region, US or EU.

  • On macOS and Linux, use the full path to the registration tool: /usr/local/bin/agenticsecurity-install on macOS, and /usr/sbin/agenticsecurity-install on Linux.

agenticsecurity-install isn't found on Windows

Open a new administrator terminal after installing the .msi, so the updated system path is loaded.

Fix an invalid Team ID

OS

What to do

macOS

Correct the Team ID in the profile or settings file, then reinstall the package.

Windows

Correct TEAMID, then run the install again.

Linux

Correct /etc/agentsecurity/enroll.conf, then run sudo dpkg --configure -a.

Uninstalling the package doesn't remove the agent

Uninstalling the installer package removes only the agenticsecurity-install command. The Agentic Security agent keeps running. To stop monitoring a device, see Remove the Agentic Security agent below.

Step 3: Turn on your policies

After the agent is installed, assign policies to decide what Agentic Security monitors. Create and assign them in Agentic Security > Policies, or from each device's Policies tab.

Policy

What it does

Finds autonomous AI agents, such as OpenClaw, Hermes, and Grok Bot, and reports how they're set up. With Log Agents set to openclaw, it also uploads OpenClaw's event logs.

Controls which tools OpenClaw can use

Optional. Records sessions and tool activity of AI coding agents: Claude Code, Claude Desktop, Codex, Cursor, Cline, and Warp.

See the List of Agentic Security Policies for every policy you can assign, including blocking policies such as AI Agent Command Protection.

The Activity Monitoring Policy collects prompts and responses. Session records can include prompts, responses, project paths, and tool results. Complete your privacy review, and tell users before you turn it on. See the policy's article for details.

Assign the inventory policy from a device

  1. Open a device enrolled in Agentic Security.

  2. Go to the Activity tab. If the device has no policy, it shows Missing Policy.

  3. Click Create/Assign Inventory Policy. The tab refreshes to show the policy's status.

Step 4: Review AI agents on your devices

Open a device in the Agentic Security device list to see:

  • Installed AI agents: OpenClaw, Hermes, Grok Bot, Claude Code, Codex, and Cursor, with their versions, and whether they're running (on Windows, OpenClaw and Hermes don't show running status).

  • Risky settings: for example, an OpenClaw gateway exposed to the network with no authentication.

  • OpenClaw events: under Logs. Events upload about every minute. The Source column shows which part of OpenClaw each event came from, and you can filter events by keyword and date.

  • AI coding agent sessions: if the Activity Monitoring Policy is on. Sessions appear within about 15 minutes. AI agent activity monitoring isn't available yet on Intel Macs or Windows ARM64. Agent inventory and the other features work.

  • Device details: encryption status, OS version, and last connected time.


Automatic agent updates

After the Agentic Security agent is installed, it updates itself on macOS, Windows, and Linux. This applies whether you deployed it with Swif MDM, another tool, or by hand. Your fleet gets the latest protections and fixes without new installers or deployment changes.

  • No redeployment: you don't need to push new installers, or change your deployment policies, when a new version is released.

  • Signed updates: updates are downloaded from Swif's servers, and are signed by Swifteam.

  • Policies keep working: your policies and CLI privilege restrictions stay in place after an update.

Check the agent version on a Mac. Run this in Terminal:

/usr/local/agentsecurity/agentsecurity --version

If a device still shows an older version, make sure it's online. Then restart it, which also restarts the agent, and check again.


CLI agent privilege restriction and root protection

To stop AI agents from changing system settings, accidentally or deliberately, Agentic Security stops supported CLI agents, such as openclaw, from running with root or sudo privileges.

Protection

What it means

Sudo and root blocking

Running monitored CLI agents with sudo, full binary paths (such as /opt/homebrew/bin/openclaw), root shells (sudo -s), or preserved environments (sudo -E) is blocked.

Normal use keeps working

CLI agents keep working with standard user permissions, so developer workflows aren't disrupted.

Only AI agents are restricted

Other admin commands, such as sudo whoami or sudo cat /etc/hosts, work as usual.

Persistent

Rules stay in place across restarts and CLI updates.

Every user

Protection applies to all non-root accounts on the device.

Logged

Blocked attempts show a permission error, and are logged on the device, in /usr/local/agentsecurity/, and on the device's Activity tab in Swif.


Remove the Agentic Security agent

To stop monitoring a device, without deleting it:

  1. Open the Agentic Security device list.

  2. Find the device, and open its action menu.

  3. Click Remove Agentic Security Agent.

  4. Confirm in the window, after checking the device name and serial number.

  5. The device's Agent Status changes to Inactive.

This removes the agent, its background service, and its data from the device, on macOS, Windows, and Linux. It works however the agent was installed, including with an installer package.

  • The device must be online. The agent is removed the next time the device connects to Swif.

  • Unassign the installer package first, if you deployed it with another MDM or Intune. Otherwise, your MDM may install the agent again.

  • Uninstalling the installer package isn't enough. It removes only the agenticsecurity-install command, and the agent keeps running.

Delete an Agentic Security device

To permanently remove a device from Agentic Security:

  1. Open the Agentic Security device list.

  2. Find the device, and open its action menu.

  3. Click Delete Device.

  4. Confirm the deletion.

  5. The device is removed from the list, and no longer appears in search or filter results.


Not an enterprise customer?

If you don't have a Swif enterprise workspace, you can still use Agentic Security with a free consumer account. See Getting Started with Agentic Security (for Consumer).


Frequently asked questions

Which AI tools does Agentic Security monitor?

Tools

Covered by

OpenClaw, Hermes, and Grok Bot

The Swif Agent Inventory Policy. It also uploads OpenClaw's event logs.

Claude Code, Claude Desktop, Codex, Cursor, Cline, and Warp

The Swif AI Agent Activity Monitoring Policy

Can I deploy Agentic Security to all my managed devices at once?

Yes. With Swif MDM, use the Agentic Security Deployment Policies on the Agentic Security page to deploy to every eligible Mac, Windows, and Linux device. Devices managed by another MDM or Intune can be deployed silently with the installer packages. See Silent deployment with another tool.

Do I need to set anything up on each device?

No. With bulk or silent deployment, each device installs the agent and enrolls automatically.

Is Agentic Security the same as the Swif MDM agent?

No. Agentic Security is a separate product, focused on monitoring and securing AI agents. It doesn't replace MDM features such as remote lock, wipe, or fleet management.

Is the setup step shared across my organization?

Yes. Once anyone in your organization clicks Download Installer, the setup step is marked complete for everyone.

Can I skip the Agentic Security setup step?

Yes. The Agentic Security step in the Setup Guide is optional, and doesn't block other onboarding steps.

How is enterprise onboarding different from consumer?

Enterprise users turn on Agentic Security in their existing Swif workspace, with no separate account. Consumer users sign up on their own.

Can I use Agentic Security on BYOD devices?

Yes. Agentic Security supports company-owned and personal (BYOD) devices. The deployment policies and Agentic Security policies support BYOD.

What's the difference between removing the agent and deleting the device?

Removing the agent sets the device to Inactive, but keeps the device record. Deleting the device removes it from Agentic Security entirely.

Does uninstalling the installer package remove the agent?

No. It removes only the agenticsecurity-install command, and the agent keeps running. Use Remove Agentic Security Agent in the device list instead. If you deployed the package with another MDM or Intune, unassign it first, so it isn't installed again.

Does Agentic Security rate each action as safe or risky?

No. It flags risky settings, such as an exposed OpenClaw gateway, but doesn't rate individual events. Review events under Logs, and filter them by keyword and date.

Where did the Agentic Security Logging Policy go?

It's been replaced by the Swif Agent Inventory Policy. OpenClaw logging is now its Log Agents setting. Existing setups keep working, with no action needed.

Can developers run CLI agents like OpenClaw with root or sudo?

No. Agentic Security blocks CLI agents from running with sudo or in root shells, to prevent privilege escalation and unintended system changes. Developers can keep running them as standard users.

Does blocking sudo for AI agents affect other admin commands?

No. The restriction applies only to monitored AI CLI tools. All other sudo commands keep working.

Where can admins see blocked privilege escalation attempts?

Blocked sudo attempts are logged with times and usernames on the device, in /usr/local/agentsecurity/, and shown on the device's Activity tab in Swif.

Do I need to redeploy Agentic Security when a new version is released?

No. Enrolled agents update themselves on macOS, Windows, and Linux. You only need to deploy again to devices where the agent was removed or never installed. The Agentic Security Monitoring Gap Detected alert on the Agentic Security page shows which devices those are.

Do automatic updates affect BYOD devices?

Yes. Automatic updates work the same way on company-owned and BYOD devices.


Need help?

If you run into any issues during setup, contact Swif support at support@swif.ai.

Related resources

Did this answer your question?