Claude for Chrome

How to Automate your Entire Workflow using Claude (10 Copy & Paste Prompts)

Jun 03, 2026

You’re switching tabs 47 times a day.

Open analytics. Copy numbers. Paste into a doc. Open Gmail. Search for a thread. Open Calendar. Check the agenda. Back to Gmail.

Rinse. Repeat. All morning.

That’s not work. That’s the administration of work. And it’s eating 2–3 hours every single day.

Claude for Chrome just changed that. Not “it might help” changed it. Actually, materially, eliminating it changed it.

I spent the last two weeks testing it across every workflow I could think of, analytics, inbox, competitor research, content creation, client management, the whole thing. Some automations worked on the first try. Some didn’t. I’ll tell you which is which.

First: What Claude for Chrome actually is.

It’s not a chatbot in a sidebar. That’s the mistake everyone makes.

Claude for Chrome is a browser agent. It can see your tabs. Click buttons. Fill forms. Scroll pages. Extract data from behind logins. Navigate between sites. And now, schedule recurring tasks.

The old way: Log in → scrape data manually → copy → paste → format → summarize → share.

The new way: Open tabs → describe what you want → Claude does it → you review and approve.

If this is useful, share it with one person who still spends Monday mornings in a tab-switching spiral. It takes 10 seconds.

Before you run anything — read this.

Two settings matter.

First: Turn on “Ask before acting.” Every time Claude wants to click something, fill a form, or take an action on a new site, it will pause and ask you first. Always start here. You can loosen the leash once you trust it.

Second: Don’t run it on financial or banking tabs yet. Claude for Chrome is in beta. Anthropic has been upfront about prompt injection risks: malicious instructions hidden in web pages that can hijack the agent’s behavior. It’s rare, but it’s real. Stick to analytics, Gmail, Drive, and content sites for now.

Third: Pair it with Cowork for deliverables. Chrome is for gathering data and taking actions on the web. Cowork (in the Claude Desktop app) is for turning that data into finished files, Excel models, PowerPoint decks, and Word reports. They work seamlessly together.

Now. The 10 workflows.

Workflow 1: Summarize Any Webpage in Seconds

Time saved: 5–15 minutes every time you need to read something you don’t have time to read.

This is the easiest place to start. No setup. No dashboards. No logins. Just open any page- an article, a landing page, a long blog post, a product page, and ask Claude to make sense of it for you.

Most people new to Claude for Chrome try to start with something complex - wrong move. Start here. Get comfortable with how it sees and reads web pages. Then level up.

Example 1 - Understand a pricing page before a sales call:

You’re about to jump on a call with a SaaS vendor. Their pricing page has 4 tiers, 30 feature checkboxes, and an asterisk on everything.

Open their pricing page and run this:

I have a SaaS pricing page open in this tab.

Read it and tell me:
1. How many tiers are there, and what are their prices?
2. What's the main difference between the cheapest and most expensive plan?
3. What features are locked behind the higher tiers that most small teams actually need?
4. Is there a free trial or money-back guarantee mentioned anywhere?
5. What's the one thing about this pricing I should ask the sales rep about?

Plain text. Under 150 words.

Workflow 2: The Inbox Triage + Promotional Cleanup

Time saved: A Saturday afternoon compressed into 15 minutes.

You have 14,000 unread emails. Most is noise. You’ve been meaning to fix it for two years. You never do because doing it manually feels like a full day of punishment.

Open Gmail and run this:

I have Gmail open. Scan my inbox (Primary, Promotions, and Updates tabs)
for the last 90 days. Classify each sender into:

1. Promotional: Marketing, newsletters, sales blasts, coupon codes
2. Automated: Receipts, shipping updates, calendar notifications,
   order confirmations, password-change alerts
3. Signal: Emails from real people, or automated emails I likely need
   to keep (invoices, tax docs, legal notices)

Then:
- Build a list of the top 20 senders in buckets 1 and 2.
- Show me: sender name, email count in last 90 days, most recent date.
- For each sender, give one recommendation:
  "Unsubscribe and delete all" / "Keep but archive old" / "Review manually."

Do NOT delete or unsubscribe from anything yet. Wait for my approval.

Once I approve senders, mark all emails from them and move to Trash in batches.
For unsubscribes: click the unsubscribe link in their most recent email.
Pause before any action that sends an email on my behalf or changes account settings.

Workflow 3: The Competitor Research Deck (Chrome → Cowork Handoff)

Time saved: 3 days compressed into 2 hours.

Your VP wants a competitive landscape deck by Friday. Five competitors. Pricing, features, positioning, and recent updates. You know how this goes, tab hell for three days, then a Sunday assembling slides.

Open your five competitor URLs and run this:

Research these 5 competitors and build me a comparison deck:
[URL 1], [URL 2], [URL 3], [URL 4], [URL 5]

For each competitor, visit their site and extract:
- Tagline/one-line positioning (from homepage hero)
- Pricing tiers (plan name, monthly price, top 3 features per tier)
- Top 5 features as listed on their features or product page
- Target customer (SMB, mid-market, enterprise — infer from the site)
- Most recent blog post or changelog (title, date, one-sentence summary)
- Any "why us" or comparison page they have

Save everything as structured JSON in a single artifact.

If a competitor hides pricing behind "Contact Sales," mark it as Custom.
Do not guess pricing. Pause for my review before the deck build.

Then hand off to Cowork in Claude Desktop and ask it to build a PowerPoint:
- Slide 1: Title slide "Competitive Landscape: [my product]"
- Slide 2: Positioning matrix (price vs. target customer size)
- Slides 3–7: One slide per competitor — logo, tagline, pricing table,
  top features, "what they do well / where they're weak"
- Slide 8: Side-by-side pricing comparison table
- Slide 9: Feature parity matrix (rows = features, columns = competitors)
- Slide 10: 3 recommendations on where we should differentiate

Workflow 4: The Evening Meeting Prep

Time saved: 10 minutes per meeting, every day, forever.

You walk into a meeting cold. You half-remember the last conversation. Someone asks you a specific question, and you wish you’d read the thread. This costs you professionally more than you realize.

Open Google Calendar and Gmail, then run this:

I have Google Calendar and Gmail open.
For every meeting on my calendar tomorrow between 8 am and 6 pm:

1. Read the event title, attendees, description, and any linked Google Docs.
2. If there's a linked agenda doc, pull the most recent 5 bullet points
   or action items.
3. Search my Gmail for the last 07 days of threads with each external attendee.
   Summarize the latest unresolved item from each thread in one line.
4. If I have any unsent drafts to those attendees, flag them.

Then build a single "Tomorrow's Prep" page with one block per meeting:
- Meeting title, time, attendees
- Context in 2 lines: what this meeting is for
- Last we left it: most recent decision or open question from email or the doc
- What I should bring up: 2 specific things where I owe someone an answer
- Open drafts: any unsent emails to attendees

Format as a clean page I can read in under 4 minutes.
Schedule this to run automatically every weekday at 6 pm.

Workflow 5: The Google Drive Cleanup

Time saved: An entire Sunday afternoon.

“Untitled document.” “Copy of Copy of Q3 plan.” “Final FINAL v3.” Every Drive has these. They multiply. You ignore them. They take up storage and mental space.

Open Google Drive and run this:

Open my Google Drive. Scan everything in "My Drive" (not Shared with me).

1. Identify all files named "Untitled document," "Untitled spreadsheet," etc.
   For each, open it, read the first 200 words, and suggest a real title
   based on the content.

2. Identify duplicates — same content, near-identical names like
   "Q3 plan" and "Copy of Q3 plan." Keep the most recently edited.
   Propose moving the others to a folder called "_duplicates."

3. Group all loose files at the root level into folders by inferred project.
   Use existing folder names where they fit.
   Propose new folders only when nothing fits.

4. Flag any document I haven't opened in 12+ months as an archive candidate.
   Don't move them yet — just list them.

Show me the full proposed plan as a single review page before touching any file.
I will approve in batches.

Workflow 6: The LinkedIn Content Research Sprint

Time saved: 1–2 hours of manual research before every content session.

Before writing a LinkedIn post, you need to know what’s already performing. What angles are working? What your audience just shared. Most people skip this step and wonder why their posts don’t land.

Open LinkedIn and run this:

I have LinkedIn open. I want to write a post about [TOPIC].

Search LinkedIn for posts on this topic published in the last 7 days.
For each of the top 10 posts by engagement:

- Copy the first line (the hook)
- Record the approximate like/comment count
- Identify the content format (list, story, opinion, data drop)
- Note the one thing that makes the hook work

Then give me:
1. The 3 hook patterns that appeared most across high-performing posts
2. The 2 angles I haven't seen yet that could stand out
3. A suggested hook for my post using the best-performing pattern

Do not write my full post. Just give me the hook analysis and one suggested hook.

Workflow 7: The Job Application Research Pack

Time saved: 2–3 hours per application.

Before any interview or application, you should know the company inside out — their latest news, their values page word-for-word, and recent LinkedIn posts from the hiring manager. Almost nobody does this. The ones who do get the job.

Open the company website and LinkedIn, then run this:

I'm applying to [ROLE] at [COMPANY]. I have their website and LinkedIn open.

Visit both and pull:

1. From their website:
   - Mission statement (exact words from their About page)
   - Top 3 values listed (exact wording, not paraphrase)
   - Any recent announcements, press releases, or blog posts in the last 90 days
   - Their main product lines or services, one sentence each

2. From their LinkedIn:
   - Last 5 company posts (title, date, topic)
   - Any posts from the hiring manager or department head (last 30 days)
   - Employee count and recent growth trend (if shown)

3. Build a 1-page "Company Intel Brief" with:
   - Company snapshot (3 lines)
   - What they're focused on right now (from recent posts)
   - 3 questions I should ask in the interview (based on what you found)
   - 2 values-alignment lines I can weave into my answers

Format as a clean brief I can read in 5 minutes before the call.

Workflow 8: The Client Onboarding Intake

Time saved: 30–60 minutes of back-and-forth intake calls.

Before a client call, you have to gather context — their website, their social profiles, their recent content, and their competitors. You usually do this the night before in a panic. This workflow does it automatically the moment they book.

Open the client’s website, LinkedIn, and any other URLs they’ve shared:

I'm onboarding a new client: [CLIENT NAME] at [COMPANY URL].

Visit their website and LinkedIn. Build a client context brief:

From their website:
- What they sell (one sentence)
- Who they sell to (target customer, inferred)
- How they position themselves (tagline + one line from their hero)
- Any current offers, sales, or campaigns running

From their LinkedIn:
- Last 5 posts (topic + format + engagement estimate)
- What content themes do they repeat most
- Any obvious gaps (topics their competitors cover that they don't)

From their 3 main competitors (I'll paste URLs below):
- How each competitor positions against my client
- Any feature or offer that my client is missing

Compile into a 1-page Client Brief that I can open 5 minutes before the call.
Include 3 "quick win" recommendations I can lead with in the first conversation.

[PASTE COMPETITOR URLS HERE]

Before you move on, slow down here.
Most people will read all 10 of these, feel inspired, and run zero of them.
Here’s what I want you to do instead: pick ONE. The one that, if you ran it tomorrow morning, would give you back the most time. Set up Claude for Chrome right now. Run it on that single workflow. That’s it.
The compounding effect of saving 45 minutes every Monday is 39 hours a year. That’s a full work week. Back in your pocket.

Workflow 9: The Proposal Research Sprint

Time saved: 2–4 hours per client proposal.

Before you write a proposal, you need to understand the prospect’s world better than they do. Their recent news. Their team structure. Their competitors. Their pain points based on what they’ve been posting about. Almost no one does this research. The ones who close more deals.

Open the prospect’s website, LinkedIn, and any social platforms they’re active on:

I'm writing a proposal for [PROSPECT COMPANY]. Their website is [URL].
They're a [company type] that sells to [target market].

Visit their website and any social profiles I have open. Then:

1. Identify the 3 biggest challenges they're likely facing right now
   (based on their content, news, and job listings — infer from what they're
   hiring for and what they're talking about)

2. Find their latest funding round, acquisition, product launch, or company news
   (check their blog, press page, and LinkedIn — last 6 months only)

3. Pull 3 quotes or statements from their leadership or company content
   that reveal their strategic priorities

4. Identify their top 2 competitors and note one thing about each competitor
   does better than them right now

5. Build a "Proposal Intelligence Brief" with:
   - Company context (3 lines)
   - The problem they're most likely trying to solve (your best inference)
   - 2 lines I can open the proposal with that show I understand their world
   - The single strongest reason they should hire me over anyone else
     (based on what you've found)

Format as a clean brief, under one page.

Workflow 10: The Weekly Content Performance Debrief

Time saved: 1 hour of manual reporting every Friday.

Most creators post. Check likes. Move on. That’s not a strategy. A real content operation looks at what worked, why it worked, and what to double down on next week. Nobody has time to do this manually. Now you don’t have to.

Open LinkedIn Analytics, X Analytics, and any other platforms you’re active on:

I have my LinkedIn Analytics and X/Twitter Analytics open in tabs.

For each platform, pull the last 7 days of content performance:

LinkedIn:
- Top 3 posts by impressions (post snippet, impressions, reactions, comments)
- Top 3 posts by engagement rate (same fields)
- Total impressions vs. previous 7 days
- Follower growth vs. previous 7 days

X/Twitter:
- Top 3 tweets by impressions
- Top 3 tweets by engagement (likes + replies + reposts)
- Total impressions vs. previous 7 days
- New followers vs. previous 7 days

Then analyze:
1. What content format (list, story, image, thread, poll) drove the most impressions?
2. What topic drove the most engagement?
3. Was there one post that dramatically outperformed everything else? If yes, why?
4. What's the one content type I should post more of next week based on this data?

Output as a "Weekly Content Debrief" — one page, clean, ready to share with a team.
Include a "Next Week's Bet" section: the one format + topic combination I should lead with.

If this issue is changing how you think about browser automation, share it with one person building a content or creator business. It’ll save them days.

Next issue: I’m building a full AI content operation using only Claude tools - Chrome for research, Cowork for production, and Claude.ai for strategy. I’ll show you the exact setup, what broke, and what I’d do differently. Don’t miss it.

Here are a few issues that you might have missed (go check them out):

Thanks for reading AI in public! This post is public, so feel free to share it.

Become a paid subscriber: Get every past issue of AI In Public, plus exclusive resources, prompts, and early access to everything I build next.

And for shorter takes between issues, follow me on X → Hamza Khalid

Hamza 💙

PS: Which of these 10 workflows are you going to run first, and what’s the task you’ve been manually doing every week that you’re most ready to hand off?

Hit reply - I read every response.

AI in public

Discussion about this post

Ready for more?