How to Use Claude Co-work's Built-In Browser for Real Work
Claude Co-work now ships a built-in browser that can audit subscriptions, pull analytics, and research products. Here's how it works.
What is Claude Co-work’s built-in browser?
Claude Co-work now has a browser built directly into the desktop app, meaning Claude can navigate websites, click through interfaces, log into accounts, and pull data back into your conversation without you manually copying anything. It’s the same idea Chat GPT and Grok bot rolled out versions of in the same week, part of a broader shift toward AI tools that can act on the open web instead of just answering questions about it. You turn it on in Claude desktop under Settings, then Co-work, by switching the preferred browser mode away from “Claude in Chrome” to the new built-in option.
TL;DR
- Claude Co-work added a built-in browser that runs inside the desktop app, letting Claude click, scroll, and log into sites on its own instead of relying on a Chrome extension.
- This is different from Chat GPT’s cloud browser, which runs on remote infrastructure so tasks can keep going even when your own computer is off, and different again from Grok bot’s full cloud computer, which includes a desktop, file system, and browser together.
- Demonstrated use cases include auditing app subscriptions for wasted spend, pulling analytics data straight from a dashboard like YouTube Studio, and researching product alternatives by actually browsing retailer sites instead of skimming search results.
- The feature can be paired with scheduled tasks, so a subscription audit or analytics report can run automatically once a week without you starting the conversation each time.
- Access has been inconsistent across accounts, rolling out unevenly regardless of paid plan or region, so absence of the option in your settings doesn’t necessarily mean anything is broken.
- Claude also unified memory between the Co-work and standard chat interfaces this week, and reorganized the claude.ai interface so projects and auto-execution mode are separate, clearer buttons.
Other agents start typing. Remy starts asking.
Scoping, trade-offs, edge cases — the real work. Before a line of code.
How is this different from a Chrome extension or web search?
AI browsing has moved through a few distinct stages, and it’s easy to conflate them.
The first stage was plain web search, where a model queries the internet and summarizes what it finds, without ever touching a live page. The second stage, which arrived roughly a year or two ago, was a Chrome extension that let tools like Claude or Chat GPT remote-control your actual browser: moving a virtual mouse, clicking buttons, filling in fields on sites you’re already logged into. That approach works but is clunky, because most websites are built for human hands, not robotic ones.
In parallel, developers pushed the Model Context Protocol (MCP), a standard that lets AI tools connect directly into a service’s backend without touching the website at all. MCP is cleaner when it’s available, but most everyday services, banking portals, niche SaaS dashboards, community forums, don’t offer it. For those, a browser is still the only way in.
Claude Co-work’s built-in browser is a step past the extension model: it’s a browser embedded in the app itself rather than one riding on top of your Chrome install. It still runs on your own machine and requires the Claude desktop app to be open. Chat GPT’s new cloud browser goes a step further, running on remote infrastructure so a task can continue even if your laptop is closed. Grok bot goes furthest, offering not just a browser but a full cloud computer, complete with a file system and other applications, essentially a machine you’re renting that operates independently of your own hardware.
What can you actually do with it?
A handful of concrete use cases stand out from early testing.
Subscription auditing. Give Claude access to accounts like a text-to-speech service or an image generation tool, and it can browse into your usage and billing pages, then tell you whether the subscription is worth keeping based on how much you’re actually using it. Because the browser is built in, this kind of check can be scheduled to run on a recurring basis, say weekly, instead of something you have to remember to do yourself.
Analytics pulling. Traditionally, getting data out of a dashboard like YouTube Studio meant navigating the site yourself, exporting a CSV, and uploading it into a chat window before any analysis could start. With a built-in browser, Claude can log into the dashboard directly, pull the relevant numbers, and turn them into a finished chart or summary in one pass. Once it has login access, you can ask it to keep surfacing specific reports on a schedule, or simply ask what reports would be most useful to you given your goals, and let it decide what’s worth digging into.
Remy doesn't build the plumbing. It inherits it.
Other agents wire up auth, databases, models, and integrations from scratch every time you ask them to build something.
Remy ships with all of it from MindStudio — so every cycle goes into the app you actually want.
Community and forum research. If you belong to gated communities, private forums, or platforms like Reddit, Claude can browse into those spaces (assuming you’re a member and it has access) and summarize what other people are doing, discussing, or building. Given enough context about your own work, it can look for people tackling similar problems or surface ideas worth adapting.
Working around platforms that resist scraping. Some social platforms actively block automated access, which is usually why third-party scraping tools and subscriptions exist. Testing found that Claude’s built-in browser could still get through in at least one case, on X, producing a report from a list of accounts without any extra scraping service. Results here will vary by platform and are likely to change as sites adjust their defenses.
Product research. Given an image of a product, such as a shoe, and a prompt asking for comparable alternatives, Claude’s browser went out and actually clicked through retailer sites rather than relying on search result snippets. That deeper crawl reportedly surfaced more options than a standard search would, though it took roughly 20 minutes to complete, a reminder that browser-based research trades speed for depth.
Is it worth turning on?
For anyone already paying for Claude Co-work, turning on the built-in browser costs nothing extra and mainly requires a settings change. The tradeoff is patience: browsing tasks take longer than a text answer because the model is genuinely navigating pages step by step, not pattern-matching on cached knowledge. It’s best suited to jobs that would otherwise involve manual exporting, repeated logins, or checking several dashboards by hand, especially ones you’d want repeated on a schedule.
It’s less useful if you need an instant answer, or if the site in question already has a clean MCP integration or API, since those routes tend to be faster and more reliable than clicking through a UI. Access has also been spotty. Some users report the option isn’t showing up in their Claude settings yet even though the rollout is supposed to be underway, and that seems unrelated to plan tier or region. If you don’t see it, checking back after some time is the only real fix right now.
What else changed in Claude recently?
Alongside the browser feature, Claude unified memory across the Co-work and standard chat interfaces, so context built up in one no longer stays siloed from the other. The claude.ai interface was also reorganized: projects and the auto-execution mode used to be crowded into a single area and now sit as separate, distinct buttons, along with account-level memories organized into categories you can browse and edit directly.
Frequently Asked Questions
What is Claude Co-work?
Claude Co-work is Anthropic’s workspace-oriented version of Claude, built for running tasks like browsing, research, and project work rather than just holding a chat conversation. It now includes a built-in browser, in-app memory, and project organization tools.
Do I need the Chrome extension to use Claude’s new browser?
No. The new built-in browser replaces the need for the Chrome extension approach. You enable it in Claude desktop under Settings, then Co-work, by switching the browser mode setting.
How is Claude’s browser different from Chat GPT’s cloud browser?
Remy doesn't write the code. It manages the agents who do.
Remy runs the project. The specialists do the work. You work with the PM, not the implementers.
Claude’s built-in browser still runs through the Claude desktop app on your own machine. Chat GPT’s cloud browser runs on remote infrastructure, so tasks can keep running even after you close your laptop, putting it a step further along the same spectrum.
Can I schedule tasks to run automatically with Claude’s browser?
Yes, based on demonstrated use cases like subscription audits and analytics reports, tasks that use the browser can be set up to run on a recurring schedule rather than triggered manually each time.
Why don’t I see the browser option in my Claude settings?
Rollout has reportedly been inconsistent, with the feature appearing for some accounts and not others regardless of paid plan or geographic location. If it’s missing, waiting and checking back is currently the only known workaround.