browser-use

Automate browser navigation, form filling, and data extraction via Playwright MCP.

9|2|Updated Jan 31, 2026
One-click install
npx skills add https://github.com/AbdullahMalik17/Hacathan_5 --skill browser-use-abdullahmalik17
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: browser-use
Source: https://github.com/AbdullahMalik17/Hacathan_5/tree/main/.claude/skills/browser-use
Command: npx skills add https://github.com/AbdullahMalik17/Hacathan_5 --skill browser-use-abdullahmalik17

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @playwright/mcp, and includes scripts (resource) and references (resource) components.

What problem does it solve? Manually browsing websites, filling forms, and scraping page data is repetitive and error-prone. This Skill lets an AI agent drive a real browser through the Playwright MCP server to perform web interactions programmatically. ## Core Features & Use Cases - Full Browser Control: Navigate pages, click elements, type text, fill forms, select dropdowns, handle dialogs, and manage tabs using accessibility snapshot element references. - Data Extraction & Inspection: Capture accessibility snapshots, take screenshots, run JavaScript on the page, and inspect console messages and network requests. - Server Lifecycle Management: Helper scripts start and stop the Playwright MCP server with a shared browser context so state persists across calls. - Use Case: Automatically log into a web portal, fill out a multi-field support form, submit it, wait for the confirmation message, and save a screenshot of the result. ## Quick Start Start the Playwright MCP server with the start-server script, then navigate to a website and take a snapshot to see the interactive elements.

Frequently Asked Questions about browser-use

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate browser actions with Playwright MCP?▼

Start the Playwright MCP server on port 8808 with the shared-browser-context flag, then call browser tools through the mcp-client.py script. Use browser_navigate to open pages and browser_snapshot to get element references for clicking and typing.

How do I fill out and submit a web form programmatically?▼

Take a browser_snapshot to get element refs, then use browser_fill_form with the field refs and values, or browser_type with the submit option set to true. Finish by clicking the submit button and waiting for confirmation text with browser_wait_for.

Why does the browser lose state between MCP calls?▼

State is lost when the Playwright MCP server runs without the --shared-browser-context flag, since each call then gets a fresh browser context. Start the server with that flag to maintain cookies, tabs, and page state across multiple tool calls.

What should I do when a browser click fails or an element is not found?▼

Run browser_snapshot first to refresh the current element references, since refs become stale after page changes. If a click still fails, try browser_hover before clicking, or use browser_run_code to execute Playwright code directly.

Can I run custom Playwright code instead of individual tool calls?▼

Yes, the browser_run_code tool accepts a JavaScript function that receives the Playwright page object. This is recommended for complex multi-step operations that should execute atomically in a single call.