browser automationby SkillsIndex

Best MCP Servers for Web Browsing and Browser Automation in 2026

From headless Chromium to stealth scrapers, we ranked every browser automation MCP server in our index. The winner has 26,000 GitHub stars and a perfect 100/100 score.

Giving your AI agent the ability to browse the web is one of the highest-leverage capabilities you can add to any workflow. Research, data extraction, form filling, monitoring — all become possible with a single MCP server install.

We indexed and scored every browser automation tool in our catalog. Here are the eight worth knowing.

The Top 8 Browser Automation MCP Servers

1. Chrome DevTools MCP — Score: 100/100

GitHub stars: 26,401 · Category: Browser Automation

Chrome DevTools MCP earned a perfect score in our index — the only browser tool to do so. It exposes Chrome DevTools Protocol directly to your agent: DOM inspection, network request interception, JavaScript execution, screenshot capture, and performance profiling. Built for coding agents that need to debug and interact with real browser environments, not just scrape content.

Best for: debugging, automated testing, agents that need to interact with JavaScript-heavy SPAs

Note: requires a running Chrome instance. Not a headless-first tool.

2. Puppeteer MCP Server — Score: 94/100

GitHub stars: 1,870 · Category: Browser Automation

The Puppeteer MCP Server wraps Google's Puppeteer library as an MCP interface. Navigate URLs, click elements, fill forms, take screenshots, extract content — all via natural language commands to your agent. The Puppeteer foundation means it handles anti-bot measures better than simpler HTTP-based scrapers.

Best for: form automation, login flows, JavaScript-rendered content

Install: npx @modelcontextprotocol/server-puppeteer

3. Firecrawl MCP Server (Official) — Score: 91/100

GitHub stars: 5,560 · Category: Browser Automation / Search

The Firecrawl MCP Server combines web scraping with search — a single server that can crawl pages, extract structured content, and search the web. The output is clean Markdown, which is far more token-efficient than raw HTML. Official server maintained by the Firecrawl team.

Best for: research agents, content extraction pipelines, websites where LLM-ready output matters

Requires: Firecrawl API key (has a free tier)

4. OpenBrowser — Score: 91/100

GitHub stars: 8,934 · Category: Browser Automation

OpenBrowser is an autonomous toolkit for browser-based agents. It goes beyond page scraping — the agent can plan multi-step browsing tasks ("research competitors, summarize findings, draft a report") without step-by-step instructions. 8.9k stars suggests strong community validation.

Best for: autonomous research agents, multi-step browser workflows

5. Playwright Skill (Claude) — Score: 86/100

GitHub stars: 1,767 · Category: Browser Automation

The Playwright Skill brings Microsoft's Playwright framework to Claude as a skill. Playwright's advantage over Puppeteer: it supports Chromium, Firefox, and WebKit — useful for cross-browser compatibility testing. The skill form factor (no separate server process) makes it simpler to integrate into Claude-native workflows.

Best for: cross-browser testing, agents already using Claude skills

6. Browser Tools MCP — Score: 76/100

GitHub stars: 7,080 · Category: AI-LLM

Browser Tools MCP focuses on one thing: surfacing browser console logs directly in Cursor, VS Code, and other MCP-compatible clients. 7k stars reflects how common this need is — debugging a web app with an AI assistant that can't see the console has always been frustrating. This fixes it.

Best for: web development, debugging, live console monitoring

7. BrowserWing — Score: 85/100

GitHub stars: 752 · Category: Browser Automation

BrowserWing takes a novel approach: record your browser actions once, then convert them into reusable MCP commands. No scripting required — you interact with the browser normally and BrowserWing captures the flow. Useful for automating repetitive browser tasks without writing Playwright code.

Best for: non-developers, repetitive form filling, recorded automation flows

8. Stealth Browser MCP — Score: 85/100

GitHub stars: 336 · Category: Browser Automation

Stealth Browser MCP addresses a specific problem: scraping sites with aggressive anti-bot systems. It rotates fingerprints, handles CAPTCHAs, and spoofs browser signals. Newer tool (336 stars) but growing quickly for a reason — most other scrapers fail on Cloudflare-protected sites.

Best for: sites with Cloudflare, PerimeterX, or other bot detection

Comparison Table

ToolScoreStarsBest use caseRequires API key?
Chrome DevTools MCP10026.4kDebug + interact with live pagesNo
Puppeteer MCP Server941.9kForm automation, JS-heavy sitesNo
Firecrawl MCP915.6kResearch + clean content extractionYes (free tier)
OpenBrowser918.9kAutonomous multi-step researchNo
Playwright Skill861.8kCross-browser testingNo
Browser Tools MCP767.1kConsole log monitoring in IDENo
BrowserWing85752Recorded automation flowsNo
Stealth Browser MCP85336Anti-bot bypassNo

Browse all browser automation tools →

mcp-serversbrowser-automationplaywrightpuppeteerweb-scraping

Enjoyed this?

Get the next issue of The Weekly Index delivered to your inbox every Thursday.