Edit

Local AI Agent Baseline - OpenClaw security baseline settings reference for Microsoft Intune

This article is a reference for the settings that are available in the Local AI Agent Baseline - OpenClaw security baseline for Microsoft Intune.

This baseline limits the use of unauthorized local AI agents such as OpenClaw by configuring device settings that disrupt commonly used execution paths. Included firewall rules restrict outbound network communication from common local agent runtime environments like Node.js.

Important

These settings might not fully block all agent execution paths. This baseline includes controls that restrict runtime environments (for example, Windows Subsystem for Linux and Node.js) which can be leveraged by local agents. This baseline might also block other processes in addition to OpenClaw. Review and test each setting before deployment, and disable settings that have an unacceptable impact on legitimate workloads.

Tip

To identify devices that have local AI agents installed before deploying this baseline, use the properties catalog to collect Local AI Agent inventory data.

About this reference article

Each security baseline is a group of preconfigured Windows settings that help you apply and enforce granular security settings that the relevant security teams recommend. You can also customize each baseline you deploy to enforce only those settings and values you require. When you create a security baseline profile in Intune, you're creating a template that consists of multiple device configuration settings.

This article displays:

  • A list of each setting with its configuration as found in the default instance of that baseline version.
  • When available, a link to the underlying configuration service provider (CSP) documentation or other related content from the relevant product group that provides context and possibly additional details for a settings use.

When a new version of a baseline becomes available, it replaces the previous version. Profile instances that were created before the availability of a new version:

  • Become read-only. You can continue to use those profiles but can't edit them to change their configuration.
  • Can be updated to the current version. After you update a profile to the current baseline version, you can edit the profile to modify settings.

To learn more about using security baselines, see:

Local AI Agent Baseline - OpenClaw (Preview), Version 1

Windows Subsystem For Linux

  • Allow WSL1
    Baseline default: Enabled
    Learn more

  • Allow the Windows Subsystem For Linux
    Baseline default: Enabled
    Learn more

Firewall

  • Firewall Rule Name
    Baseline default: Configured
    Learn more

    This baseline includes two preconfigured firewall rules. Both rules block outbound TCP connections from Node.js executables to disrupt common execution paths used by OpenClaw.

    Rule: block nodejs in LOCALAPPDATA folder

    Property Default value
    Enabled Enabled
    Name block nodejs in LOCALAPPDATA folder
    Interface Types All
    File Path %LOCALAPPDATA%\Programs\node\node.exe
    Network Types FW_PROFILE_TYPE_ALL
    Direction The rule applies to outbound traffic
    Action Block
    Protocol Configured - 6 (TCP)

    Rule: block nodejs in ProgramFiles folder

    Property Default value
    Enabled Enabled
    Name block nodejs in ProgramFiles folder
    Interface Types All
    File Path %ProgramFiles%\nodejs\node.exe
    Network Types FW_PROFILE_TYPE_ALL
    Direction The rule applies to outbound traffic
    Action Block
    Protocol Configured - 6 (TCP)

    For details about firewall rule properties, see Firewall CSP - FirewallRules.