This is Claude Code controlling my browser. Am I scared? No. Am I impressed? Yes.
Clicks. Forms. Researching. I’m not touching anything. Claude is doing all of it.
I’m Charles Dove, and on the Charlie Automates YouTube channel (@charlieautomates) I show you what’s actually possible with AI automation. This one blew my mind. And the setup takes less than five minutes.
The 30,000 Foot View
Before we get into use cases, let’s break down what this system is and what you need.
Three main components:
- Claude Code + VS Code - The brain and the hands
- The MCP Server (Playwright) - The bridge
- Your Browser - The canvas
Claude is the intelligence. It decides what to click, what to type, where to navigate. The MCP server is the connection between Claude and the browser. And the browser? That’s where Claude does whatever you need it to do online.
Think of it like hiring a virtual assistant that can see your screen and take action. Except this one doesn’t need sleep, doesn’t miss steps, and doesn’t forget instructions.
What You Need to Get Started
The setup is dead simple.
You’ll need an environment. Whether you’re using Cursor, regular terminal, or VS Code like me. If you don’t have Claude Code set up yet, I have a video linked above that walks you through the full install inside VS Code. Go watch that, then come back.
Once you’re ready, grab the Playwright MCP link from the video description. Copy it and paste it into Claude Code with a prompt like:
“Set up this MCP server and list it as an available MCP server in this workspace.”
That’s it. Claude handles the rest.
Quick Tip Before You Start
When the browser opens, it won’t have you logged into your normal Chrome profile. It’s a fresh session. So log into whatever accounts you need before you start giving Claude tasks.
This trips up a lot of people. Don’t skip it.
Use Case 1: Market Research to a Spreadsheet
This is where I started. A simple request. I asked Claude to look up the top local AI automation agencies, pull up their websites, identify their top three services, and put everything into a spreadsheet.
That’s it. One prompt. Claude opened the browser, ran the Google search, visited each site, extracted the info, and organized it.
Here’s what makes this powerful. You could ask Claude to do the research AND fill in the spreadsheet for you. No copy-pasting. No switching between tabs manually. It just does it.
And the real kicker? While Claude is running this research in the browser, you can open another terminal and use Claude for something else. You can literally minimize the browser window and go work on other things.
It’s a background agent. Running while you do real work.
Maybe you’re on a sales call and need research done on a prospect. Maybe you want competitor analysis before a pitch. Maybe you’re building a lead list. Claude handles it while you focus on high-value tasks.
At CC Strategic, we use this exact workflow for competitive research before client calls. It saves hours every week.
Use Case 2: DM Automation
This one is my favorite.
I stacked some skills and custom agent frameworks into Claude, then told it to answer all the DMs in my LinkedIn inbox. I gave it context about what services I offer. I plugged in a DM response framework. Then I let it go.
Claude opened LinkedIn. Read the messages. Crafted personalized responses based on my framework. And sent them.
While it’s answering DMs, I can go do whatever I want. Build a workflow. Take a call. Record a video. Whatever.
The key here is the stacking. You can combine:
- Custom agent frameworks (like a DM response template)
- Multiple MCP servers (Playwright for browsing, n8n for workflows)
- Multiple terminal sessions (each running a different task)
So picture this. One terminal is handling your DM responses. Another terminal is building n8n workflows. A third is doing market research. All running at the same time. All autonomous.
That’s not a dream. That’s what I’m doing right now.
Use Case 3: CRM and Business Dashboard Analysis
The third use case is a CRM. An admin dashboard.
I had Playwright come in and look at my conversations. My opportunities. My pipeline. I had it analyze my workflows and pull key insights about what I should be focusing on in my business.
Think about that for a second. A personal business consultant that logs into your CRM, reads your data, and tells you where the bottlenecks are.
Maybe you have an Airtable with your numbers. Or a spreadsheet tracking your business metrics. Claude can look through it, find patterns, and tell you where to focus.
You can have a personal business consultant that comes in and analyzes your business’s performance. There’s a million ways to use this.
Sure, there are dedicated MCP servers for specific CRMs and dashboards. But Playwright gives you a universal key. If it has a browser interface, Claude can interact with it.
Stacking MCP Servers (The Real Power Move)
Here’s where things get wild.
An MCP server is just a toolbox. And you can stack as many toolboxes as you want.
I could be running Playwright for browser automation on one terminal. My n8n workflow building MCP on another terminal. And maybe a file system MCP on a third.
Multiple terminals. Multiple MCP servers. All running in parallel. All doing real work while you focus on the things that actually need your brain.
This is what true AI automation looks like for business owners. Not replacing you. Multiplying you.
What This Is NOT (Stay Grounded)
I have to be real with you. This isn’t magic.
This is a tool that lets you output more. Do more with the time you have. It’s leverage. Pure and simple.
If you think you’re going to use Anthropic’s Claude and Microsoft’s Playwright MCP server to run your business without any setup or oversight, this isn’t the video for you.
You still need to:
- Set up the system properly
- Give Claude clear instructions
- Review the output
- Stay involved in your business
AI will not replace you as the front runner for your business. That’s your job. This just handles the repetitive stuff so you can focus on what matters.
When NOT to Use Playwright MCP
I’d highly recommend not using this for:
- Simple API requests - If there’s an API, use the API. It’s faster and cheaper.
- High frequency scraping - Playwright isn’t built for that. You’ll burn tokens and hit rate limits.
- Anything with a native endpoint - If the tool has an MCP server or API, use that instead.
You might say, “Charlie, there are better use cases. The ones you described could be done with API keys.” That’s fair. But the fact that you now have this in your arsenal is the biggest win.
Playwright MCP shines when there’s no API. When the only way to interact with something is through a browser. That’s where this tool becomes irreplaceable.
If You’re Only Using Claude for Coding, You’re Missing Out
This is something I say on Charlie Automates all the time. If you’re only using Claude Code for writing code, you’re leaving so much power on the table.
Claude Code + MCP servers turn your terminal into a command center. Browser automation. Workflow building. Research. Outreach. CRM analysis. All from the same place.
This is just one of many MCP servers I’m going to cover on the channel. I’m excited to produce more videos around Claude Code, MCP servers, and efficient workflows that make your life easier and actually make you money.
Step-by-Step Setup (Quick Recap)
For those who want the fast version:
- Open VS Code with Claude Code installed (or your preferred terminal)
- Copy the Playwright MCP link from the GitHub repo
- Paste it into Claude Code and ask: “Set up this MCP server and list it as an available MCP server in this workspace”
- Wait for Claude to configure everything
- Ask Claude to open a browser to verify it works
- Log into your accounts in the new browser session
- Start prompting with real tasks
That’s the full setup. Five minutes tops.
Three Prompts to Try Right Now
Once you’re set up, try these:
Market Research Prompt:
“Open the browser, search for the top 5 [your niche] agencies in [your city], visit each site, list their top 3 services, and organize everything in a spreadsheet.”
DM Response Prompt:
“Open LinkedIn, go to my messages, read each unread conversation, and draft professional responses using this framework: [paste your response guidelines].”
CRM Analysis Prompt:
“Open [your CRM URL], navigate to my pipeline, screenshot each stage, and give me a summary of where deals are stuck and what I should focus on this week.”
Start simple. Get comfortable. Then stack.
FAQ
What is the Playwright MCP server?
It’s an open-source tool from Microsoft that lets Claude Code control a web browser. Claude can click buttons, fill forms, navigate pages, read content, and take screenshots. All through natural language prompts.
Do I need to know how to code to use this?
No. You just need Claude Code installed and a basic understanding of how to write prompts. The setup takes one prompt. Using it takes plain English.
Is this free?
The Playwright MCP server is free and open source. You’ll need a Claude Code subscription from Anthropic to use Claude Code itself. But the browser automation part costs nothing extra.
Can Claude see everything on my screen?
No. Claude only sees what’s happening inside the Playwright browser window it opens. It can’t see your regular Chrome browser, your desktop, or any other apps. It’s isolated.
Will this get my accounts banned?
It depends on how you use it. Don’t use it for high-frequency scraping or violating a platform’s terms of service. For normal browsing, research, and occasional messaging, you should be fine. Use common sense.
Can I run multiple browser sessions at once?
Yes. You can open multiple terminal sessions in VS Code, each with its own Claude Code instance and Playwright browser. They run independently. This is how you stack tasks.
How many tokens does browser automation use?
More than regular Claude Code tasks. Every page Claude reads, every screenshot it takes, every action it performs costs tokens. Keep your prompts focused and your tasks specific to manage costs.
What’s the difference between Playwright MCP and other browser automation tools?
Playwright MCP is designed specifically to work with AI models like Claude. It’s not just automation. It’s AI-driven automation. Claude decides what to do, adapts to what it sees, and handles unexpected situations. Traditional browser automation scripts break when a page changes. Claude adjusts.
Can I use this with Cursor instead of VS Code?
Yes. Playwright MCP works with any environment that supports Claude Code and MCP servers. Cursor, VS Code, terminal. Your choice.
Is this safe for business use?
Yes, as long as you review outputs before sending anything externally. I’d recommend starting with read-only tasks (research, analysis) before moving to write tasks (sending messages, filling forms). Always supervise the first few runs of any new workflow.
Ready to Go Deeper?
This is just one tool in the stack. If you want to see how I combine Claude Code, MCP servers, n8n workflows, and custom agent frameworks to run an entire business, there are two ways to get involved:
Want hands-on guidance? Work with me 1-on-1 and I’ll help you build your own AI automation stack.
Want to learn with a community? Join CC Strategic AI on Skool where I share free Claude Code templates, workflows, and weekly breakdowns of exactly what’s working.
Subscribe to @charlieautomates on YouTube for more tutorials like this. I’m covering more MCP servers, Claude Code workflows, and AI automation strategies throughout 2026.
The tools are here. The question is whether you’ll use them.