SAN FRANCISCO–(BUSINESS WIRE)–Claude Cowork puts an autonomous agent in every employee’s hands, not just developers. Anyone in the company can now connect Claude to Slack, Gmail, HubSpot, their file system, and dozens of other tools, then hand it a task and let it run. With Anthropic bringing Cowork to web and mobile, it’s no longer confined to the desktop. Sessions now run in Anthropic’s cloud, follow employees across devices, and keep working even when no device is online.
That employee in finance who just connected Cowork to the shared drive and their email? They didn’t file a ticket or ask security. They just clicked “connect”.
Today we’re announcing Opsin support for Claude Cowork, giving security and governance teams the visibility and guardrails they need to let their organization use Cowork safely.
What Is Claude Cowork?
Cowork is Anthropic’s agent for non-developers. An employee describes what they want done, and Claude does the work: reading and writing files, searching email, updating CRM records, posting to Slack, pulling data from internal systems through connectors.
Where Claude Enterprise chat answers questions, Cowork takes actions. It reads real files, calls real APIs, and moves real data between real systems. And it does this on behalf of employees across every department, each running with their own permissions and their own set of connected tools.
Cowork started on the desktop. As of August 3, it also runs on web and mobile, with sessions hosted in Anthropic’s cloud. An employee can kick off a task from their phone, pick it up on their laptop, and let scheduled tasks run in the background around the clock. Adoption just got easier, which means usage is about to grow, and so is the surface your team is responsible for governing.
Why Claude Cowork Matters for Security Teams
Every Cowork session is a chain of tool calls the security team never sees. An agent might read a folder of contracts, summarize them, and send that summary through a connector to an external service. Each step looks reasonable on its own. Together, they can move sensitive data somewhere it was never supposed to go.
The risks compound in ways existing tools weren’t built to catch:
- Employees connect Cowork to data sources on their own, with no security review and no central record of what’s connected where
- A single prompt can trigger dozens of tool calls, each one touching files, email, or business systems
- An overshared folder or an overly permissive connector becomes an active exposure the moment an agent reads from it
- The same connector that’s approved for one team’s workflow may violate policy when another team uses it with regulated data
- With web and mobile access, sessions run in the cloud and continue when no one is watching, so risky activity is no longer bounded by a managed laptop or a workday
DLP and CASB tools watch data moving through predictable channels. An agent chaining tool calls across email, files, and SaaS apps in a single session is not a predictable channel.
The answer isn’t to block Cowork. Employees adopt these tools because they work. The answer is to see what’s happening and govern it, so the organization gets the productivity without the exposure.
Claude Cowork Governance with Opsin
Opsin now covers Claude Cowork as part of our AI governance platform, alongside Claude Enterprise, Claude Managed Agents, Microsoft Copilot, ChatGPT Enterprise, and Google Gemini.
→ Full Visibility Into Cowork Activity
Opsin gives security teams a complete picture of Cowork usage across the organization: who is running sessions, what they’re asking for, and every tool call the agent makes along the way. Coverage spans desktop, web, and mobile, so it doesn’t matter where a session starts or which device picks it up. Chats and actions are captured with full context, so when a question comes up about what an agent did and why, the answer is a lookup, not an investigation.
→ Alerts on Risky Actions
Not every tool call is equal. An agent summarizing a public doc is routine. An agent reading a folder of employee compensation data and sending it through an external connector is not. Opsin analyzes tool calls in the context of the data they touch and alerts your team when an action puts enterprise sensitive data at risk, before a routine session becomes an incident you read about in a postmortem.
→ Connector Governance
Connectors are where Cowork’s power and its risk both live. Opsin inventories every connector in use across your Cowork deployment, shows which employees and sessions are using them, and checks that usage against your organization’s governance policy. When a connector is used outside its approved purpose, or when someone connects a tool that was never approved at all, your team knows.
Enable Cowork, Don’t Fight It
Security teams have been here before. A new tool spreads through the business faster than governance can keep up, and the choice looks like block it or accept the risk. Opsin gives you a third option: let employees use Cowork, with the visibility to know what’s happening and the guardrails to keep sensitive data where it belongs.
That’s what safe enablement looks like. Your organization gets the productivity gains, and your security team gets the control. Nobody has to choose.
See Your Full Claude Risk Picture with Opsin
Opsin’s Claude coverage now spans Chat, Projects, Managed Agents, and Cowork: every way Claude operates in your enterprise, governed from one platform.
About Opsin
Enterprise AI (Microsoft Copilot, ChatGPT, Claude, Gemini) is the new endpoint. The enterprise workforce is already using it, creating agents faster than security teams can track. That opens a new attack surface: data oversharing, insider risk, and agent sprawl that traditional tools weren’t built to secure. Opsin gives security teams the visibility and control to secure enterprise AI, with 1-click deployment via API. Our proactive risk assessment simulates real user queries to surface AI risk within 24 hours, so security teams can move quickly and confidently to adopt AI.
Opsin understands user and agent behavior through our context graph – connecting identity, data, and model to deliver the accuracy teams need to prioritize what actually matters, and remediate at the root cause, not just triage alerts. Trusted by leading enterprise teams.
Contacts
Media Contact:
Caroline Maier
VP Marketing, Opsin
[email protected] | 917.324.0160
