StackJack 0.1.0.13267+608a1a7
StackJack 0.1.0.13267+608a1a7
This release adds client organizations to AI Governance, a Windows endpoint agent for discovering and controlling AI tools on devices, and new sources for Gemini, Vertex AI, Amazon Bedrock and a range of security products. It also fixes activity pulls that could stop progressing on busy time periods.
New features
- Client organizations. You can now set up a separate organization for each client inside your AI Governance account. Each client organization has its own:
- Sources and Microsoft connection. The client's own administrator can grant Microsoft consent.
- Staff directory, DLP rules and jurisdiction.
- Audit view.
- Evidence packs. Privacy and auditor packs name the client organization and count only that client's activity.
An organization picker in the Portal switches every page to the selected organization, and a new Organizations tab lists your client organizations with a detail page for each.
- Client administrators. You can invite people to a single client organization and give them roles that apply only there. They can work with the following for their own client only:
- The alert inbox
- Alert emails
- Teams approval cards
- Closing a client organization. A closing client organization shows the date its data will be removed, and you can reopen it until that date. After that, its data is removed, including:
- Enrolled devices
- Enrollment keys
- Its agent branding
Tenant-wide administrators can still open a read-only view of closed organizations using the "Show closed" switch on the organizations list. Switching off a Gemini conversations source, or closing its organization, leaves the StackJack matter in your Google Vault.
- Organization-aware governance tools. The AI governance tools now accept an organization, so an agent can work against a specific client. A new tool lists the organizations you can see. If a name matches more than one organization, the tool asks for the organization's ID instead.
- Endpoint agent for Windows. A new agent, installed from the Endpoint agent page, enrolls devices into your MSP organization or a client organization using an enrollment key. Each key is shown only once. On each device, the agent:
- Discovers installed AI apps, AI browser extensions, running AI processes, visited AI domains, local MCP servers and installed security products.
- Blocks unapproved AI apps, or warns the user and then allows them.
- Blocks unapproved AI domains.
- Checks pastes into AI tools.
- Applies tool policy to local MCP servers that don't go through StackJack.
- Works with Claude Code, Cursor and the Gemini CLI.
- Shows a tray icon and on-screen notices to users.
- Updates itself. You can hold devices on a specific version.
If the agent can't reach StackJack, it allows activity unless your policy is set to block in that case.
- Uninstall code. You can require a code to uninstall the endpoint agent. The code must be 12 to 64 characters. Upgrading or replacing the agent package does not ask for the code.
- Security products handle enforcement when connected. When an endpoint agent capability is set to automatic and one of these products is connected, the product handles that capability:
- Microsoft Defender with Purview
- DNSFilter
- Cisco Umbrella
- ThreatLocker
- SentinelOne
Devices that don't run the product are still covered by the agent.
- White-label agent packages. You can set up a brand for your own endpoint agent package and a separate brand for each client organization from the new brands page.
- Endpoint agent alerts. Two new alerts tell you, with a device count:
- When devices are running an outdated agent version.
- When devices have gone silent or show signs of tampering.
The Endpoint agent page shows coverage counts.
- Browser extension fallback. The browser extension can now send its events through the endpoint agent when it cannot reach StackJack directly.
- Security product sources. You can now connect the following products to report AI-related activity they see:
- Microsoft Defender (alerts and advanced hunting)
- DNSFilter
- Cisco Umbrella
- ThreatLocker
- SentinelOne (threats and STAR alerts)
- CrowdStrike Falcon
- Huntress (AI misuse incident reports and app control signals)
- Bitdefender GravityZone
- ESET PROTECT
- Google Gemini sources.
- Gemini activity. Reads Gemini activity from Google Workspace and matches it to your staff directory.
- Gemini conversations. Reads conversations through Google Vault exports. These sources are created switched off, and StackJack asks you to confirm before turning one on. A domain larger than Google's search limit can add one source per organizational unit. If the Vault matter is closed, you can start a new one from the health page.
- Cloud AI sources. A new Cloud AI page lets you add:
- Google Vertex AI request and response logs. These are read from your own logging table using a service account key you provide.
- Amazon Bedrock invocation logs. The page gives you the setup steps and policies to apply in your AWS account, then asks for the role to use.
Improvements
- Data-subject exports now include the person's endpoint agent devices, with:
- Device reference
- Platform
- First and last time seen
- Revocation date, if revoked
- Erasing a staff member now removes their link to any endpoint agent devices while keeping the device records.
Bug fixes
- Cloudflare, DefensX and Island activity pulls could stop progressing when a time period contained more data than could be read in one pass. These pulls now split the period into smaller pieces. If a piece is still too large, StackJack records a visible gap and continues.
Known issues
- Branded endpoint agent packages still use the StackJack name for the Windows service, the install folder and the command-line tool in this release.