How to Give LLM Access to Your Browser (the Easy Way)

I just discovered an amazing MCP that allows LLMs to use your browser, and it's extremely simple to set up and use.

There is a lot of hype around MCP, which is absolutely deserved. The problem is that most MCP use cases are related to coding, such as giving LLMs access to databases or up-to-date documentation. However, I just discovered an amazing MCP that allows LLMs to use your browser, and it's extremely simple to set up and use.

This MCP is called BrowserMCP.

Here, I’ve connected it to Claude Desktop with Pro account, so that I can use it without any additional API costs, but you can also easily connect it to any other app that supports MCP.

It’s open source, and all automation happens locally on your machine so your browser data isn't sent to remote servers.

Now, let's see how it works!

First, I’m going to give BrowserMCP access to this browser tab.

Then, I’ll ask Claude:
Use BrowseMCP to check LinkedIn for new job openings that match my CV. I’m an AI engineer looking for B2B, worldwide and remote positions. Show me the three most relevant ones.
I’m also going to attach my CV.

And that's where the magic happens - Claude can now operate in your browser tab where you are logged into all the services you need to automate. It's not lightning fast, but it's faster than Browser Use, one of the most popular LLM browser automation solutions.

It works with Claude Desktop, so you don't need to pay for API use. As you can see, it actually works pretty well. Let's wait a bit and see what result it provides!
Claude offered me three relevant jobs. Nice! I’ll look into the details and apply.

Now, let's see how to set up BrowserMCP. It's actually extremely simple. I’m going to follow the official guide at browsermcp.io
First, open the Claude desktop settings and click on "Developer.

Then, choose "Edit Config."

Double-click the claude_desktop_config.json here and just copy and paste this code from the documentation.

Finally, install the Chrome extension, and you're all set!

Let's test it. I'm going to ask Claude to write a post for my Substack based on the script I used for this video. I'm giving you access to this browser tab. Now, I'm going to ask Claude:
Use BrowseMCP and create a draft of a post on my Substack based on the attached text. Don't post anything; just format the post nicely.

The result is impressive. It actually created the entire Substack post. The formatting isn't ideal, but this level of automation still feels natural.

I just started this Substack - let me know if you find these short, practical tutorials helpful, and I’ll share more AI tricks and techniques!