microsoft

microsoft/playwright-mcp

TypeScript Apache-2.0 AI & ML

Playwright MCP server

34.9k stars
2.9k forks
active
GitHub +279 / week

34.9k

Stars

2.9k

Forks

2

Open issues

30

Contributors

v0.0.78 09 Jul 2026

AI Analysis

Playwright MCP provides a Model Context Protocol server that enables LLMs to automate web browsers using Playwright's accessibility tree instead of screenshots. It is purpose-built for AI agents performing web automation, exploratory testing, and autonomous workflows—particularly well-suited for systems that benefit from persistent browser state and iterative reasoning over page structure, but less ideal for high-throughput coding agents where token efficiency is critical.

AI & ML Developer Tool Discovery value: 3/10
Documentation 8/10
Activity 10/10
Community 9/10
Code quality 5/10

Inferred from signals mentioned in the README (tests, CI, type safety) — not a review of the actual code.

Overall score 8/10

AI's overall editorial judgment — not an average of the bars above, can weigh other factors too.

browser-automation model-context-protocol llm-integration accessibility-tree ai-agents
Actively maintained Well documented Popular Niche/specialized use case Apache-2.0 licensed Production ready
Deep Analysis · Based on README and public signals
2w ago

Microsoft's official Playwright MCP server lets LLMs drive browsers via accessibility trees, not screenshots

playwright-mcp exposes Playwright's browser automation capabilities as a Model Context Protocol server, allowing LLMs and AI coding agents to interact with web pages through structured accessibility snapshots rather than pixel-based vision models. Built by Microsoft's Playwright team, it targets developers building agentic workflows, autonomous test generation, and exploratory web automation. Its accessibility-tree approach is token-efficient and model-agnostic, making it suitable for non-vision LLMs. It integrates with VS Code, Cursor, Claude Desktop, Copilot, Codex, and many other MCP clients out of the box.

Origin

Created March 2025 as MCP protocol adoption accelerated across the AI tooling ecosystem. Positions itself as the official Microsoft-backed bridge between LLMs and Playwright, complementing the separately maintained playwright-cli project.

Growth

Reached 34K stars in roughly 15 months — extremely rapid for a developer tooling repo. Growth is driven by the simultaneous explosion of LLM coding agents, Microsoft's brand authority in the browser automation space, and the MCP protocol becoming a de facto standard for agent tool use. The 330 stars/week rate as of June 2026 indicates sustained but moderating interest after an initial surge.

In production

Wide integration support across at least 10 named MCP clients (VS Code, Cursor, Windsurf, Claude Desktop, Copilot, Codex, Cline, Amp, etc.) suggests real adoption pressure drove those integrations. Microsoft's direct involvement adds institutional credibility. However, quantified production usage metrics (npm downloads, telemetry data, case studies) are not present in the available README. Adoption appears broad among early AI agent developers but scale is not independently verified.

Code analysis
Architecture

Likely implements an MCP server in TypeScript that wraps Playwright's Node.js API. Appears to expose browser actions (navigate, click, type, extract) as MCP tools, with accessibility tree snapshots as the primary page representation rather than screenshots. Based on README, the server maintains persistent browser context across tool calls, which distinguishes it from stateless CLI invocations.

Tests

Not documented in README

Maintenance

Last push was June 23, 2026 — one day before the evaluation date — indicating very active, ongoing maintenance. As an official Microsoft repo under the Playwright organization, it benefits from dedicated engineering resources. README explicitly addresses architectural tradeoffs between MCP and CLI approaches, suggesting thoughtful, evolving design rather than a one-shot release.

Honest verdict

ADOPT IF: you are building LLM agents or agentic workflows that need to interact with live web pages, especially where persistent browser state, accessibility-tree structured data, and cross-browser support matter. AVOID IF: you are running high-throughput coding agents with tight context-window budgets — the README itself recommends playwright-cli+SKILLS in that scenario. MONITOR IF: you are evaluating MCP as a protocol standard for agent tooling; this repo's trajectory will reflect broader MCP ecosystem health.

Independent dimensions

Mainstream potential

8/10

Technical importance

8/10

Adoption evidence

6/10

Risks
  • MCP protocol itself is still maturing; if a successor standard emerges or MCP falls out of favor, this server's value proposition changes significantly.
  • Accessibility tree fidelity varies across sites — heavily dynamic, canvas-based, or accessibility-poor applications may yield poor snapshots, degrading agent reliability.
  • Token cost of verbose accessibility trees can be prohibitive at scale; the README's own comparison against CLI+SKILLS acknowledges this tradeoff explicitly.
  • Dependency on Microsoft maintaining active interest; if Playwright organizational priorities shift, maintenance pace could slow despite current signals being strong.
  • Community MCP alternatives (e.g., mcp-chrome) may close feature gaps, reducing the differentiation of the cross-browser approach for users who only need Chrome.
Prediction

Likely to become the default browser automation layer for MCP-based agents as the ecosystem matures, especially given Microsoft's backing and the breadth of client integrations already established.

0 found this helpful

Newsletter

Get analyses like this every Monday

Free weekly digest of the most interesting open-source discoveries.

Languages

TypeScript
61.9%
JavaScript
32.2%
Dockerfile
5.8%

Information

Language
TypeScript
License
Apache-2.0
Last updated
19h ago
Created
16mo ago
Analyzed with
anthropic/claude-haiku-4-5

Stars over time

Loading…

Contributors over time

Top 100 contributors only — repos with more will plateau at 100.

Loading…

Similar repos

executeautomation

executeautomation/mcp-playwright

Playwright MCP Server is a Model Context Protocol implementation that enables...

5.6k TypeScript AI & ML
microsoft

microsoft/playwright

Playwright is a Microsoft-backed browser automation and end-to-end testing...

92.6k TypeScript Dev Tools
microsoft

microsoft/playwright-cli

Playwright CLI is a command-line interface for Playwright browser automation,...

11.9k JavaScript Dev Tools
remorses

remorses/playwriter

Playwriter is a Chrome extension and CLI tool that enables AI agents to control...

3.7k HTML Dev Tools
MicrosoftDocs

MicrosoftDocs/mcp

The Microsoft Learn MCP Server is a Model Context Protocol (MCP) integration...

1.8k TypeScript AI & ML
vs. alternatives
executeautomation/mcp-playwright

Community-built MCP Playwright integration with 5.5K stars. Predates this official package and likely influenced its creation. Now overshadowed by Microsoft's official version in terms of trust, maintenance resources, and client integration support.

hangwin/mcp-chrome

Chrome-specific MCP server with 12K stars. Focuses on Chrome DevTools Protocol rather than Playwright's cross-browser abstraction. More narrowly scoped but potentially lower overhead for Chrome-only use cases.

microsoft/playwright-cli

Microsoft's own alternative for coding agents. README explicitly positions CLI+SKILLS as more token-efficient for high-throughput coding agents, while MCP is better for persistent-state, iterative reasoning workflows. The two co-exist rather than compete directly.

microsoft/playwright

The underlying automation library. playwright-mcp is a thin MCP wrapper around it; users needing direct programmatic control without LLM orchestration should use Playwright directly.

remorses/playwriter

Smaller-scope project (3.6K stars, HTML) focused on AI-assisted Playwright script writing rather than runtime browser control via MCP. Different use case — code generation vs. live agent control.