Day 6 of 14 · Claude Cowork Challenge

Browser Automation

⏱ 6 min 📊 Medium

So far, Claude has been working with files on your computer. But a huge amount of knowledge work happens in the browser — reading articles, extracting data from websites, filling out forms, gathering research. Today you'll learn how the Claude in Chrome extension connects Cowork to the web, giving Claude the ability to navigate, read, and interact with websites on your behalf.

This is the third pillar of Cowork's power: local files, sub-agents, and now browser automation.

The Claude in Chrome extension

The browser automation capability comes through an official Chrome extension called Claude in Chrome. This extension creates a bridge between the Claude Desktop app (where Cowork runs) and your Chrome browser.

Once installed, Claude can:

Navigate to URLs — Open specific web pages, follow links, move through multi-page sites.

Read page content — Extract text, tables, and structured data from any web page. Claude understands page layout, so it can distinguish headers from body text, navigation from content, and data tables from decorative elements.

Interact with pages — Click buttons, fill form fields, select dropdown options, scroll through content. Claude can perform the same actions you would with a mouse and keyboard.

Extract structured data — Pull specific information from web pages and organize it into spreadsheets or documents. This is web scraping without writing code.

The extension works alongside Cowork — you initiate tasks from Claude Desktop, and Claude uses the browser as a tool when it needs web-based information or actions.

Knowledge Check

How does Claude access web browsing capabilities in Cowork?

A

Through the Claude in Chrome extension, which bridges Claude Desktop and the Chrome browser

B

Through a built-in browser inside the Claude Desktop app

C

By connecting to a cloud-based browser service

D

Through a separate mobile app

The Claude in Chrome extension is the bridge. It connects the Claude Desktop app (where Cowork runs) to your Chrome browser, allowing Claude to navigate, read, and interact with web pages as part of Cowork tasks.

Setting up the Chrome extension

Getting browser automation working takes just a few steps:

1. Open Chrome and navigate to the Chrome Web Store

2. Search for "Claude in Chrome" (the official Anthropic extension)

3. Click "Add to Chrome" and confirm the installation

4. Pin the extension to your toolbar for easy access

5. Connect it to Claude Desktop — the extension will detect your Claude Desktop installation and link to it

Once connected, you'll see a small indicator in Claude Desktop confirming that browser access is available. From that point on, when Claude needs to access a website as part of a Cowork task, it will use your Chrome browser automatically.

Important: Claude uses your browser session, which means it can see websites you're logged into. This is useful — it can read pages behind authentication — but be mindful of what's open in your browser when Claude is working.

🤖

Claude

online

I need to research the pricing pages of these 4 SaaS tools: Slack, Asana, Jira, and Linear. Pull their plan names, prices, and key feature differences into a spreadsheet.

I'll use browser automation to visit each pricing page and extract the data. Opening Chrome... - Visiting slack.com/pricing... reading plan tiers... done - Visiting asana.com/pricing... reading plan tiers... done - Visiting jira.atlassian.com/pricing... reading plan tiers... done - Visiting linear.app/pricing... reading plan tiers... done All data extracted. I've created **saas-pricing-comparison.xlsx** with: 📊 **Sheet 1: Plan Overview** — All plans across 4 tools with monthly and annual pricing 📊 **Sheet 2: Feature Matrix** — Key features mapped against each plan tier 📊 **Sheet 3: Per-User Cost** — Normalized per-user/month cost for easy comparison Slack and Asana have the most complex tier structures. Linear is the simplest with just two paid tiers. Want me to add any other tools to the comparison?

↻ Replay conversation

Use cases for browser automation

Browser automation unlocks a set of tasks that were previously impossible for a desktop AI agent. Here are the most practical applications:

Web research: Instead of you visiting 10 websites, copying text, and pasting it into a document, Claude visits the sites, reads the content, and synthesizes the information directly. It's research without the tab-switching.

Data extraction: Need to pull product listings, pricing data, job postings, or directory information from websites? Claude can navigate to the pages, extract the structured data, and deliver it as a clean spreadsheet. This is what companies pay for web scraping tools to do.

Form submission: Claude can fill out repetitive web forms — supplier questionnaires, registration forms, data entry portals. You provide the data, Claude handles the clicking and typing.

Competitive monitoring: Ask Claude to visit competitor websites, check for changes in pricing, new product announcements, or updated positioning. Combine this with file creation, and you get automated competitor briefs.

Content gathering: Collecting information from multiple pages for a report, a presentation, or an analysis. Claude reads each page, extracts what's relevant, and compiles it into your preferred format.

Knowledge Check

Which of the following is a practical use case for Claude's browser automation?

A

Replacing your web browser entirely

B

Hacking into protected websites

C

Automatically posting on social media accounts

D

Extracting pricing data from multiple competitor websites and compiling it into a spreadsheet

Data extraction from websites is one of the most practical browser automation use cases. Claude navigates to each page, reads the content, extracts the structured data, and compiles it — no manual copy-pasting, no web scraping scripts to write.

What Claude can and cannot do in the browser

It's important to understand the boundaries of browser automation so you set realistic expectations.

Claude can:

- Navigate to any publicly accessible URL

- Read and extract content from web pages

- Interact with page elements (click, type, scroll, select)

- Handle multi-page workflows (navigate through pagination, follow links)

- Access sites you're logged into (using your existing browser session)

- Take screenshots of pages for reference

Claude cannot:

- Solve CAPTCHAs or bypass bot-detection systems

- Access content that requires two-factor authentication prompts during the session

- Work with websites that heavily rely on JavaScript rendering that doesn't load in time

- Guarantee that every website will be readable (some sites have unusual structures)

Best practices for browser automation:

- Be specific about which URLs to visit — don't just say "research this topic"

- If a site requires login, make sure you're logged in before starting the task

- For important data, ask Claude to show you what it extracted so you can verify accuracy

- Combine browser automation with file creation — extract data from the web, then create a document from it

Knowledge Check

What should you do before asking Claude to extract data from a website that requires login?

A

Ask Claude to create an account for you

B

Give Claude your username and password

C

Make sure you're already logged in to the site in Chrome before starting the task

D

Use a different browser that doesn't require authentication

Claude uses your existing Chrome session, so it can access sites you're already logged into. You should log in before starting the task — don't share credentials with Claude. It will use the active session in your browser.

🤖

Claude

online

I need to fill out a vendor registration form on our procurement portal for 3 new suppliers. I have their details in a spreadsheet in my work folder.

I'll read the supplier details from your spreadsheet and fill out the forms. Reading suppliers.xlsx... found 3 vendors: 1. Acme Solutions — IT Services 2. BrightPath Consulting — Management Consulting 3. DataFlow Systems — Data Infrastructure Here's my plan: 1. Open the procurement portal in Chrome (make sure you're logged in) 2. For each supplier: navigate to the registration form, fill in all fields, and submit 3. Take a screenshot of each confirmation page as proof of submission 4. Create a log file with submission timestamps and confirmation numbers Shall I proceed? Make sure you're logged into the portal first.

↻ Replay conversation

Final Check

What are the three pillars of Cowork's capabilities that you've learned about so far?

A

Text generation, image creation, and code writing

B

Chat, voice, and video

C

Reading, writing, and editing

D

Local file access, sub-agent parallel execution, and browser automation

The three pillars are: (1) local file access — reading and writing files on your computer, (2) sub-agents — spawning parallel workers for independent subtasks, and (3) browser automation — navigating and interacting with websites through Chrome. Together, they make Cowork a comprehensive work agent.

🌐

Day 6 Complete

"Files on your computer plus data from the web — browser automation connects Cowork to the entire internet."

Tomorrow — Day 7

Your Cowork Foundation

We'll recap everything from Week 1, establish best practices, and prepare for plugins and advanced workflows in Week 2.

Continue to Day 7 →