Free tools Windows power users keep installed
One-click scans. No signup required.
To give Claude Code browser access, connect Microsoft Playwright MCP: install Node.js 20 or newer, then run claude mcp add playwright npx @playwright/mcp@latest in a terminal. Claude can then navigate and interact with pages through browser tools, using structured accessibility snapshots to understand what is on screen. This guide uses “Claude’s browser MCP server” to mean Playwright MCP connected to Claude Code; other browser MCP servers may use different commands and capabilities.
What Claude’s browser MCP server does
MCP, or Model Context Protocol, is an open protocol for connecting applications to tools and context. Playwright MCP is a browser automation server that lets an MCP client such as Claude Code ask a browser to navigate pages and interact with them. See Anthropic’s MCP overview and the Playwright MCP documentation.
Instead of interpreting only a screenshot, Playwright MCP gives Claude structured accessibility snapshots of page content and controls. Claude can use that information to locate elements, follow links, fill fields, and perform actions. The server is the browser-control layer; Claude remains the assistant that decides what to do based on your request and the page information it receives.
This setup is for Claude Code. The Playwright guide names Claude Code as a supported client and documents a local server launched with npx. Claude Code also supports remote MCP connections, but remote transport is not needed for the local setup here.
#1 Best Overall
What you need before installing
- Node.js 20 or newer. The Playwright setup guide lists this as a prerequisite.
- Claude Code installed and available in your terminal.
- A working internet connection for
npxto obtain the Playwright MCP package the first time it runs.
You do not need to buy a separate browser or special hardware for the documented local setup. The package tag @latest may resolve to a newer release as the project changes, so consult the official setup instructions if the command or behavior differs later.
Install Playwright MCP in Claude Code
- Open a terminal and check your Node.js version with
node --version. Confirm that it is version 20 or newer. - Register the server by running:
claude mcp add playwright npx @playwright/mcp@latest - In Claude Code, check the MCP connection with
/mcp. If the server is listed but not yet authorized, follow the prompt to approve or connect it. - Try a low-risk browser task. For example, ask: “Navigate to https://demo.playwright.dev/todomvc and add a few todo items.”
That demo is the example used in the Playwright getting-started guide. Claude should navigate to it, inspect its structure, and use browser tools to add the items. The exact interaction and output depend on the page state and your prompt.
For real websites, start with a clearly bounded request such as finding a page or checking visible text. Review any action that would submit a form, change account settings, send a message, or make a purchase before allowing it to proceed.
Choose where Claude Code stores the MCP configuration
Claude Code offers configuration scopes for different situations. The Playwright command above is a local server launched by Claude Code; it does not require you to create a shared configuration file.
- Local scope: suitable when the connection is for your own work in the current project context.
- Project scope: a project configuration can be shared through
.mcp.json. Project-scoped servers require approval before use, so teammates can review the connection. - User scope: suitable when you want a server configuration available across projects.
Use project configuration only when sharing the server setup is useful to the team. Do not put passwords, API keys, or other secrets in a shared project file. For Claude Code’s management commands and configuration scope details, see the Claude Code MCP guide.
Run the browser visibly or headlessly
The Playwright MCP getting-started guide says headed mode is the default, meaning the browser window is visible. To run without displaying a browser window, add --headless to the server arguments. The documented browser choices are Chrome, Firefox, WebKit, and Microsoft Edge; for example, a configuration can include --browser=firefox. Consult the Playwright MCP guide for the exact configuration format supported by the release you install.
Use headed mode when you need to watch the browser or diagnose a page that behaves unexpectedly. Headless mode is useful in environments without a desktop display, such as some automated or remote workflows. Headed mode without a display may fail to launch; where the deployment fits, Playwright documents an HTTP server option for that situation.
Use the standalone HTTP server when local stdio is not the right fit
The basic Claude Code command starts Playwright MCP locally. The Playwright guide also documents running a standalone HTTP server, for example:
Recommended Free Tools
npx @playwright/mcp@latest --port 8931
In that deployment shape, an MCP client connects to http://localhost:8931/mcp. This can help when the browser server and MCP client need to connect over HTTP, including some systems without a display. It is not a necessary step for the simple local setup, and the client must be configured to use the server endpoint.
At the broader Claude Code level, MCP supports local stdio as well as remote SSE and HTTP transports. Remote servers may require additional setup or authentication; Claude Code documents OAuth authorization through /mcp for servers that use it. Do not treat remote transports or OAuth as prerequisites for Playwright’s local command.
Rank #3
Configure browser behavior for a specific job
Playwright MCP documents a JSON configuration file for advanced browser options, browser context options, network rules, and timeouts. These controls matter when a page needs a particular browser, a non-default context, or a bounded wait. Check the installed version’s official guide for current option names and syntax rather than copying an unverified configuration from another release.
- Browser selection: choose among the documented Chrome, Firefox, WebKit, or Microsoft Edge values when the target requires a particular engine.
- Headless mode: pass
--headlesswhen a visible browser is unnecessary or unavailable. - Timeouts: configure bounded waits for pages or operations that load slowly; an overly short timeout can interrupt a legitimate load, while an excessively long one delays useful failure feedback.
- Network rules: use documented rules when the task needs a controlled network environment, and avoid blocking resources that the page requires to render or function.
Browser MCP versus a screenshot API
Playwright MCP is designed for interactive browser automation: Claude can inspect page structure and take actions. A screenshot API instead returns an image or PDF from a URL, which is useful when the deliverable is a capture rather than an ongoing browser interaction. ScreenshotNeo is a developer-facing screenshot API and MCP server; see ScreenshotNeo for its service overview.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteIf your goal is a static screenshot or PDF rather than Claude driving an interactive browser, ScreenshotNeo can be called with one GET request. Its MCP server also exposes screenshot and PDF tools for AI clients.
Or skip the browser setup
For a one-off website image, you can request a capture directly rather than configure a browser automation server. The example below requests a WebP capture of Stripe; replace the URL with the page you want. See the ScreenshotNeo API documentation for request options and response details.
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webp
ScreenshotNeo accepts cookie or consent banners before capture and removes more than 60 known consent platforms, newsletter popups, and chat widgets; each cleanup step can be turned off. Bot checks, blank pages, timeouts, failed loads, and cache hits cost nothing, and response headers identify the page verdict and whether the request was billed. Its MCP server provides take_screenshot, get_page_info, and capture_pdf for Claude, Cursor, and other MCP clients. The free plan includes 1,000 screenshots per month with no card; paid plans start at $5 for 3,000 screenshots.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Sign up free for ScreenshotNeo to get 1,000 screenshots a month with no card.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Troubleshoot common setup problems
Claude Code does not recognize the command
Check that Claude Code is installed and that its command is available in the terminal’s PATH. Run claude mcp list to inspect registered servers, then run the documented add command again if Playwright is missing.
The browser does not open
Headed mode is the default in the Playwright guide. If the machine has no display, configure headless mode with --headless or use the documented standalone HTTP deployment if it matches your client and environment.
The server appears but Claude cannot use it
Open /mcp in Claude Code and check whether the server needs authorization. If it was configured in a project’s .mcp.json, approval is required before the project-scoped server can be used.
A page loads incompletely or an action times out
Pages may depend on delayed scripts, network requests, or content that appears after initial navigation. Try a less demanding test page, allow a suitable timeout through the documented configuration, and make the prompt specify the page state or control Claude should wait for. Check network rules if required page resources might be blocked.
The command works on one machine but not another
Confirm Node.js 20 or newer is available where the server runs, that npx can obtain the package, and that the second machine has the same intended browser and display mode. With @latest, the package release may differ over time; check the current Playwright setup guide if a later installation behaves differently.
Use browser control carefully
A connected browser can reach pages and interact with controls, so a request should specify the intended scope. Begin on a harmless page, verify what Claude is about to change, and require review before consequential actions such as account changes, form submissions, or purchases. Use a personal or project configuration scope deliberately, and keep secrets out of shared configuration.
Frequently Asked Questions
Does Playwright MCP work with Claude Desktop?
The setup described here is for Claude Code, which the Playwright guide names as a client. Other Claude clients may have different installation steps.
Does Claude see only screenshots of a page?
No. Playwright MCP can return structured accessibility snapshots for Claude to inspect, rather than relying only on screenshots.
Can I use a different browser MCP server?
Yes, but each server may expose different tools and use different installation commands. This guide covers Microsoft Playwright MCP.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
