Anthropic has shipped a built-in browser inside Claude Cowork, its desktop-first AI agent app. Starting August 26, Claude no longer needs your browser to interact with the web. When a task involves a website, a separate browser opens inside Cowork’s side panel, and Claude navigates, reads, clicks, and fills forms on your behalf.

It is now live on Enterprise plans and rolling out over the coming week to Pro, Max, and Team users on macOS, Windows, and Linux.

What Claude Can Do With It

The capability goes beyond reading web pages. Claude can open any website in its own browser pane, interact with page elements, submit forms, pull information from live sites, and chain those actions together as part of a longer workflow — all without switching apps or giving it access to your personal browsing session.

That last point matters. This is Claude’s browser, not a window into yours. It runs completely separate from your own Chrome, Safari, or Edge. Claude cannot see your tabs, history, bookmarks, or saved passwords. The same safeguards that govern Claude in Chrome apply here, including real-time checks that compare Claude’s actions against what you actually asked it to do.

For Enterprise deployments, admins control the feature in Organization settings under Cowork, then Built-in browser. That includes turning it on or off, managing access by team, and tracking usage across the organisation.

Why This Is a Step Change for AI Agents

Most enterprise AI tools — including previous versions of Claude Cowork — could talk about web tasks or help you draft what to send, but they required you to do the actual clicking. You were still the hands.

This release removes that bottleneck. Claude can now complete multi-step web workflows in one instruction. Research a competitor, pull a pricing table from a vendor site, fill in an intake form, grab a report from a supplier portal — these are things you would previously have had to hand-hold an AI through step by step, or just do yourself.

That changes the practical value of AI assistance from “making my work faster” to “doing work while I’m doing something else.”

It also puts Anthropic directly in competition with Google’s Gemini Spark (cloud-based background agents) and Microsoft’s Copilot browser actions — both of which launched earlier this year. The difference Anthropic is emphasising is privacy: your data stays in your own environment, and there’s no ambiguity about what Claude can see.

What This Means for Business

For teams already using Claude Cowork for research, reporting, or customer workflows, the built-in browser makes those processes genuinely autonomous. You hand off a task — “find the last three quarterly reports from this supplier and summarise the key risks” — and Claude handles the web research while you focus on the decision, not the data gathering.

For businesses running Omni-style AI workflows, this kind of agentic web access is the piece that makes multi-step automation practical. Agents that can only read documents you give them are limited. Agents that can go and get the documents themselves, verify information against live sources, and update internal systems based on what they find — that is a different category of capability.

The key question for teams evaluating this is governance. Enterprise admins have controls, but any system that can interact with web forms and portals on your behalf needs clear policies around what it’s authorised to access. That planning pays dividends early. Teams that define their agent boundaries now will deploy faster and with fewer surprises than teams that figure it out under pressure.

The rollout is gradual, and Anthropic is clear that the safeguards reduce risk without eliminating it. That’s honest. The right posture is to start with lower-stakes web tasks, validate the output, and expand scope as confidence builds.

For businesses already in the Claude ecosystem, this is a significant upgrade. For those still evaluating agentic AI tools, the gap between what was possible six months ago and what is possible today just got noticeably wider.

---

Enterprise DNA helps businesses build practical AI workflows using tools like Claude Cowork, custom AI agents, and structured data systems. If you’re exploring how agentic AI fits your operations, start with a discovery call.