Commvault (Nasdaq: CVLT) is extending its endpoint-protection capabilities to selected local AI-agent data, giving IT and security teams a way to discover tools on managed laptops, protect their configurations and recover that information after a device disruption. The move targets an emerging enterprise problem: employee AI work can depend on locally stored instructions, context and workflows that conventional file recovery may not identify as a distinct operational asset.
The company said the capability will cover supported local data associated with Claude Cowork, Claude Code and OpenClaw on managed Windows and macOS devices. Teams will be able to identify installations of those tools, protect locally stored agent configurations and recover the protected information after a laptop is lost, damaged, replaced or otherwise disrupted. Commvault said discovery of OpenAI’s ChatGPT desktop application is available now; the broader discovery, protection and recovery support is expected to become generally available in coming months.
Commvault has described a direction and a delivery window, rather than an immediately available set of protections for every named AI tool. IT leaders assessing the capability will need to confirm the supported versions, local data locations, recovery boundaries and availability dates before treating it as part of an endpoint standard.
AI-assisted work may include prompt libraries, custom instructions, generated material and tool-specific context accumulated by developers, marketers and other employees. If that material exists primarily on an endpoint, a laptop incident can interrupt work even when the employee’s ordinary files are restored. Commvault’s approach is intended to add those agent-related artifacts to a protection and recovery process that IT teams can manage centrally.
The feature also places AI-tool discovery inside an endpoint-resilience workflow. The company said teams can see supported AI tools installed outside approved IT processes and investigate where they are present. That can help organizations build a more accurate inventory of AI use on managed devices, but discovery alone does not establish whether a tool is authorized, how it is configured or whether its connected data is handled appropriately. Those questions still require policy, identity controls and review by security and business owners.
For enterprise operations, the development reflects a shift in what needs to be protected. Endpoint backup has traditionally focused on files, user settings and applications. As work moves into AI tools with persistent context and custom configurations, the recovery target can include the working state that makes those tools useful. Treating that state as recoverable may reduce the manual effort of rebuilding an employee’s AI workflow after hardware replacement, although the company has not provided performance or recovery-time benchmarks.
There are limits to what the source establishes. The capability will govern every AI service, prevent unapproved use or protect information held only in a vendor’s cloud environment. Organizations should also determine whether the data captured by an endpoint-protection policy contains sensitive business information and whether retention, access and legal requirements are aligned with internal standards.
In the end, it gives IT teams a more concrete operating model for a problem that is often handled informally. Legacy endpoint programs were built around user files and installed applications, while AI adoption introduces reusable context and configurations that may be costly to recreate but easy to overlook. Commvault includes recovery and visibility as part of the control plane for those assets, shifting the discussion from allowing AI tools on laptops to managing the continuity of the work performed inside them.