MCP server
Playwright MCP server
Microsoft's Playwright MCP server lets AI drive a real browser: open pages, read them, click, type, fill forms and take screenshots.
- Official server
- Browser automation
At a glance
- Maintained by
- Microsoft
- Tools exposed
- 10 listed below
- Reviewed
- September 24, 2026
Why it matters
What it does and why it matters
Playwright is Microsoft's widely used browser automation library, and its MCP server gives an AI assistant a real browser to work with. The assistant reads each page through a structured snapshot of the page's accessibility tree rather than by guessing from a screenshot, which makes its clicks and form fills more reliable.
That opens up two kinds of work. Development teams use it to test their own web apps by describing a flow in plain English. Business teams use it for tasks on sites that have no API, such as checking a supplier portal or collecting public information.
A browser is a powerful tool to hand to an agent. It can click buttons, submit forms and run JavaScript on a page. We run it without saved logins, limit which sites it may visit and review any action that submits data. For your own apps, point it at a test environment, not production.
Web pages are untrusted content and may contain text meant to redirect an agent. Keep the task narrow and do not let page content decide what the agent does next without a check.
Use cases
Best business use cases
Test web apps
Visual checks
Portals without APIs
Reproduce bugs
Tools
Available tools
Tool names as published by the maintainer at review time. Versions change, so check the repository for the current list.
Read and write: Read and write: reads page content and can click, type, submit forms and run JavaScript in a real browser.
-
browser_navigateGo to a URL.
-
browser_snapshotCapture an accessibility snapshot of the current page.
-
browser_clickClick an element on the page.
-
browser_typeType text into an editable element.
-
browser_fill_formFill several form fields at once.
-
browser_take_screenshotTake a screenshot of the current page.
-
browser_evaluateRun JavaScript on the page or an element.
-
browser_network_requestsList network requests since the page loaded.
-
browser_tabsList, create, close or select browser tabs.
-
browser_wait_forWait for text to appear or disappear, or for a set time.
Workflows
Example workflows
Check a signup flow after a release
-
1
Point the server at the staging site.
-
2
Ask the assistant to sign up as a new test user.
-
3
It fills the form and follows the confirmation steps.
-
4
It reports errors and takes screenshots.
-
5
A developer reviews the findings.
Read an order status from a portal
-
1
Give the assistant a portal page and a reference number.
-
2
It moves through and searches the portal.
-
3
It reads the status from the page.
-
4
It returns the status and a screenshot.
-
5
No data is submitted except the search.
Prompts
Example prompts
“Open our staging site, sign up as a new user and tell me where the flow is confusing.”
“Take screenshots of the home, pricing and contact pages at mobile width.”
“Search this supplier portal for order 48213 and tell me its status.”
Setup
Installation and configuration
Examples for Claude Desktop and Claude Code. Replace the placeholder values with your own keys, and store secrets outside version control.
Claude Desktop: claude_desktop_config.json
{
"mcpServers": {
"playwright": {
"command": "npx",
"args": [
"@playwright/mcp@latest"
]
}
}
}Claude Code: terminal
claude mcp add playwright npx @playwright/mcp@latest- Local server: stdio; @playwright/mcp; mcr.microsoft.com/playwright/mcp.
Security
Security considerations
-
Limit sites
Restrict which sites the browser may open where possible.
-
No saved logins
Do not give the agent sessions it does not need.
-
Untrusted pages
Treat page text as data, not instructions.
-
Review submissions
Require approval before the agent submits forms that matter.
Related servers
Servers that pair well with it
Browserbase
Browserbase's hosted MCP server gives AI a cloud browser that can browse, act on pages, observe elements and extract data.
Firecrawl
Firecrawl's official MCP server lets AI scrape, map, crawl and search the web and get clean, AI-ready content back.
Fetch
The MCP project's reference Fetch server lets AI download a web page and read it as clean text or markdown.
Keep exploring
Solutions and services that use it
Related services
View all related services- MCP integration Connect AI assistants to your CRM, files, databases and tools through Model Context Protocol servers, with safe permissions.
- AI agents AI that takes actions in your systems, such as qualifying leads or processing requests, with people checking the results.
- Website speed optimization Find what makes a site slow and fix it, so pages load quickly on phones and rank better in search.
- API development Well-documented APIs that let your product talk to apps, partners and AI tools reliably.
- Progressive web apps Web apps that install on a phone, work offline and send notifications, without an app store.
Solutions
View all solutions- Price monitoring Collect competitor prices automatically, match them to your products and alert you when something changes.
- AI sales research agent Short, sourced briefings on each prospect before a call, drafted by an AI agent from public information and your CRM.
- AI lead qualification Score, research and route every new inquiry so your team talks to the best leads first.
- Email triage Sort a shared inbox by topic and urgency, pull out the key details and draft replies for a person to send.
Industries
View all industriesGuides and articles
View all guides and articles- What are AI agents and how do businesses use them? What AI agents are, real business uses, the controls they need and how to start with one safely.
- What is the Model Context Protocol? A business guide to MCP: what it is, how MCP servers work, who supports it and how to start using it.
- How to build an AI workflow with MCP servers Plan, connect and test an AI workflow that uses MCP servers, from a single task to a reliable process.
- AI chatbot vs AI agent What separates a chatbot that answers from an agent that acts, and when you need each.
AI models
View all ai models- Claude Fable 5.1 The top tier of the Claude family, for the most demanding reasoning and long-horizon agent work, priced above Opus.
- Claude Opus 5.5 Anthropic's recommended starting point for most serious work: long-running agentic coding and knowledge work, with a 1M token context window.
- GPT-6 Astra OpenAI's most capable model, built for the hardest end-to-end work: complex reasoning, coding, computer use and research.
- DeepSeek V4 Pro DeepSeek's model for demanding reasoning and agent work, with thinking effort levels, tool calling, JSON output and a 1 million token window.
- Grok 4.7 xAI's current flagship for coding, agent tasks and knowledge work, with image input, tool calling, structured outputs, reasoning effort levels and a 500K window.
- Reasoning Models that think through multi-step problems before answering: analysis, planning, math, complex documents and agent work.
Glossary terms
View all glossary terms- AI agent An AI agent is software that uses a language model to plan steps, call tools and act toward a goal, instead of only answering one question.
- Progressive web app A progressive web app, or PWA, is a website that works like an installed app, with offline use, home screen icons and notifications.
- MCP server An MCP server is a program or hosted service that exposes a system's data and actions as tools that AI assistants can use through the Model Context Protocol.
- Model Context Protocol The Model Context Protocol, or MCP, is an open standard for connecting AI assistants to external tools and data in a consistent way.
- Tool calling Tool calling is when an AI model asks the application to run a tool, such as a search or database lookup, and then uses the result in its answer.
- CRM A CRM, or customer relationship management system, stores contacts, companies, deals and interactions so a team can manage its customer relationships in one place.
FAQ
Questions people ask us
Have a question that is not here? Ask us directly.
Yes. It is maintained by Microsoft, the maker of Playwright.
It mainly uses structured page snapshots, and can also take screenshots.
The Puppeteer reference server has been archived. Playwright is the maintained option we recommend.