DEV.co
MCP Servers · microsoft

playwright-mcp

Playwright MCP is a Model Context Protocol server that enables LLMs to interact with web pages using accessibility data instead of screenshots. Built and maintained by Microsoft, it provides deterministic browser automation optimized for agentic workflows that require persistent state and rich page introspection.

Source: GitHub — github.com/microsoft/playwright-mcp
34.8k
GitHub stars
2.9k
Forks
TypeScript
Primary language
Apache-2.0
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
Repositorymicrosoft/playwright-mcp
Ownermicrosoft
Primary languageTypeScript
LicenseApache-2.0 — OSI-approved
Stars34.8k
Forks2.9k
Open issues2
Latest releasev0.0.77 (2026-06-29)
Last updated2026-06-29
Sourcehttps://github.com/microsoft/playwright-mcp

What playwright-mcp is

TypeScript-based MCP server exposing Playwright's accessibility tree to LLM agents over stdio. Replaces vision-model dependencies with structured data, integrates with 15+ MCP clients (Claude Desktop, VS Code, Cursor, etc.), and requires Node.js 18+. Apache 2.0 licensed and actively maintained.

Quickstart

Get the playwright-mcp source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/microsoft/playwright-mcp.gitcd playwright-mcp# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Exploratory web automation and testing

Self-healing test agents and autonomous workflows that benefit from persistent browser context and iterative reasoning over page structure without token overhead of screenshots.

Long-running agentic loops

Agents performing complex multi-step workflows across web applications where maintaining continuous context and rich introspection justifies MCP's overhead compared to CLI-based approaches.

LLM-native browser interaction without vision models

Applications where avoiding vision model dependencies and deterministic tool application are critical—data entry, form filling, web scraping, and accessibility-driven interactions.

Implementation considerations

  • Node.js 18+ is mandatory; verify runtime availability before deployment.
  • MCP client must be compatible (VS Code, Cursor, Claude Desktop, etc.); standard config is `npx @playwright/mcp@latest`, but client setup varies slightly.
  • Tool schemas and accessibility trees are verbose; monitor token consumption in high-volume agentic loops to ensure context window remains viable.
  • Playwright manages browser instances; ensure host has adequate memory and no browser process limits for long-running agents.
  • State persists across calls; agents must handle browser navigation, session timeout, and context cleanup explicitly.

When to avoid it — and what to weigh

  • High-throughput coding agents with token constraints — Microsoft's own guidance notes CLI + SKILLs is more token-efficient for modern coding agents; MCP's verbose tool schemas and accessibility trees may exceed context budgets.
  • Simple, single-step browser tasks — If your workload is lightweight or one-off (login, fetch data, close), the overhead of MCP server setup and state management outweighs benefits compared to direct Playwright CLI.
  • Visual or pixel-level interaction required — MCP trades screenshots for accessibility trees; if your use case requires visual element detection or coordinate-based clicks, accessibility data alone will not suffice.
  • Environments without Node.js 18+ — Hard runtime requirement; cannot be used in isolated or older runtime sandboxes without meeting the Node.js version floor.

License & commercial use

Apache License 2.0 (Apache-2.0). Permissive OSI-approved license allowing commercial use, modification, and distribution with standard attribution and liability disclaimers.

Apache 2.0 is permissive and typically allows commercial use. However, verify compliance with your organization's legal review—particularly if bundling with proprietary code or if vendor support/indemnification is required beyond standard open-source terms.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityClear
Deployment complexityLow
DEV.co fitGood
Assessment confidenceHigh
Security considerations

MCP server accepts browser automation commands over stdio from the connected LLM client; no explicit input validation or sandboxing described in available data. Runs with host user permissions; browser instances can access host network and file system. No rate limiting, authentication, or encryption specified. Requires trust in the MCP client and host isolation. Consult Playwright's security posture (Not specified in provided data) and evaluate risk of arbitrary browser automation in your threat model.

Alternatives to consider

Playwright CLI + SKILLs

Microsoft's own recommendation for modern coding agents; more token-efficient for high-throughput workflows. Trade-off: loses persistent context and rich introspection that MCP provides.

Selenium Grid or Puppeteer MCP servers

Language-agnostic alternatives or Node.js-native competitors. Selenium is heavier but more mature; Puppeteer is lighter but has smaller ecosystem.

Direct Playwright API (Node.js/Python bindings)

If you control the agent code directly, skip MCP protocol overhead and call Playwright directly. Best for tightly integrated systems without LLM agent abstraction.

Software development agency

Build on playwright-mcp with DEV.co software developers

Install Playwright MCP in your MCP client and start building autonomous workflows. Review security and token costs for your use case before production deployment.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

playwright-mcp FAQ

Does Playwright MCP require vision models?
No. MCP uses Playwright's accessibility tree (structured data) instead of screenshots, eliminating vision model dependencies and reducing inference latency.
Can I use this in production for mission-critical workflows?
Early-stage versioning (0.0.77) suggests API stability is not yet guaranteed. Consider pinning a specific version, monitoring for breaking changes in minor releases, and testing thoroughly before production rollout.
How does MCP token usage compare to CLI-based Playwright?
MCP loads full tool schemas and verbose accessibility trees into context, increasing token overhead. CLI + SKILLs invocations are more concise. MCP is justified if persistent state and rich introspection outweigh token cost.
What if my MCP client is not in the supported list?
Playwright MCP implements the MCP protocol; any client that supports MCP stdio servers and can execute `npx @playwright/mcp@latest` should work. Documentation covers 15+ tested clients; others may require manual config.

Work with a software development agency

DEV.co helps companies turn open-source tools like playwright-mcp into production software. Our software development services cover the full lifecycle — architecture, web development, integration, and maintenance — delivered by software developers and web developers who ship. Engage our software development agency to implement or customize it for your mcp servers stack.

Ready to add AI-driven browser automation?

Install Playwright MCP in your MCP client and start building autonomous workflows. Review security and token costs for your use case before production deployment.