Neo brings endpoint visibility inside Claude Cowork

Key takeaways:

• Neo’s integration with Claude Cowork extends the Neo sensor into Cowork VMs, the virtual machines where Cowork runs its work.
• Security teams can see activity inside the VM, tied to the endpoint, user and Cowork session it came from.
• Cowork activity appears in the same Neo console as the rest of the device, with no separate tool to run.

Claude Cowork lets people hand an AI agent a multi-step task and come back to finished work. The task might be sorting a folder of contracts, pulling figures out of a stack of PDFs, or turning research notes into a draft report. To keep that work isolated, Cowork runs its most sensitive actions inside a virtual machine on the user’s own computer.

Isolation is good design, but it leaves security teams with a gap. Tools on the host can see that the VM is running, but not what the agent does inside it: which files it opens, which processes it starts, which hosts it contacts. As more organizations roll out Cowork, agentic work happens somewhere most security stacks can’t look.

Neo’s integration with Claude Cowork extends the Neo sensor into that VM, so security teams see Cowork activity with the same context they already have for the rest of the endpoint.

What’s new

Neo’s integration with Claude Cowork extends the Neo sensor into Cowork VMs and reports what it sees in the Neo console, tied to the device, user and session the activity came from.

Highlights include:

• Visibility inside the VM. Processes, file activity and network connections inside the Cowork VM appear in Neo, along with the AI components the agent uses.
• Session-level attribution. Cowork keeps each session separate inside the VM, and Neo ties activity to the session, user and endpoint behind it.
• The same console as the rest of the device. Cowork activity shows up next to what Neo already reports for macOS, Windows, Linux and WSL.

How it works

When a Cowork VM starts, the Neo sensor on the endpoint extends into it. That covers every Cowork session on the machine, and activity from the VM reports to the same Neo backend as the rest of the device. Admins turn the integration on with the device management tools they already use, by distributing Neo’s package and setting a Claude Desktop policy.

Coverage inside the VM builds on Neo’s Linux sensor, which shares code with the Windows and macOS sensors and already runs inside WSL. Inventory and events from the VM behave the way they do everywhere else Neo runs.

How other approaches handle Claude Cowork

Network tools see connections leaving the device, which helps spot unexpected destinations. They can’t see which file the agent read inside the VM before it made the call. MDM confirms that Claude Desktop is installed and configured, and stops there. Audit logs and APIs record what happened in a service after the fact, not live activity on the device.

EDR running inside the VM sees processes and files for threat detection. It treats Cowork like any other Linux workload, though. It lacks the AI context Neo adds: which agent components are present, and how Cowork activity relates to the other AI tools on the same laptop.

Neo runs on the device and now extends into Cowork’s VM as well. That lets it answer “what AI is running here, and what is it doing?” for the whole endpoint, not everything except Cowork.

Bottom line for teams using Claude Cowork

With Neo’s Claude Cowork integration, security teams can:

• Know what AI is running and what it can reach, including inside the Cowork VM.
• Answer “who or what did this?” for Cowork activity, down to the session.
• Investigate faster, with Cowork events in the same console as every other AI tool on the device.
• Clear security review for Cowork sooner, because the VM is no longer a blind spot.

Get started

Neo’s integration with Claude Cowork is coming soon for Neo customers using Cowork on Claude Enterprise.

To turn it on, contact your Neo account team, and read more on Claude support.

Want to see Neo’s coverage for Claude today? Join our live webinar to learn how Neo helps secure Claude across different environments on November 30 at 11:00 AM CT. The 25-minute session walks through where Claude runs across an organization, from Claude.ai and Claude Code to Cowork, and shows how Neo gives security teams visibility into each. Register now.

Frequently asked questions

Does the integration change how Claude Cowork works?

No. Neo observes and reports on activity inside the Cowork VM. Cowork itself works exactly as it would without Neo.

Does this replace the EDR we already run?

No. EDR still handles threat detection across your endpoints. Neo adds AI context that EDR doesn’t have, such as which agent components are present and how Cowork activity relates to the other AI tools on the device.

Do I need to deploy anything new to cover Cowork?

No new product. Coverage extends the Neo sensor already on your endpoints, and admins turn it on through the device management tools they use today.