remorses/playwriter

What it solves

Playwriter allows AI agents to control a user's existing, running Chrome browser instance rather than spawning a fresh, isolated browser. This solves the problem of agents lacking access to existing login sessions, cookies, extensions, and browser state, and prevents them from being instantly flagged by bot detectors that typically target fresh automation instances.

How it works

Playwriter uses a Chrome extension and a local WebSocket server to bridge the gap between an AI agent (via MCP or CLI) and the browser. Instead of launching a new process, it connects to the browser's internal debugger API. It provides a full Playwright API for automation, allowing agents to execute JavaScript code directly in the browser context.

Who it’s for

Developers building AI agents that need to interact with the web as the user does, including accessing authenticated sites and bypassing bot detection, as well as developers who want to use the Model Context Protocol (MCP) for browser automation.

Highlights

• Session Persistence: Uses your actual Chrome profile, meaning agents have access to your logins and cookies.
• ** keluarga Playwright API**: Provides the full power of the Playwright automation library rather than a limited set of tools.
• Visual Labels: Generates color-coded accessibility snapshots (aria-ref selectors) to help AI agents identify and interact with page elements more efficiently.
• Low-Latency Execution: Executes commands directly via a relay server, avoiding the overhead of spawning sub-agents or heavy screenshot-based context.
• Advanced Debugging: Includes capabilities for live code editing, network interception, and setting breakpoints within the browser.
• Remote Access: Supports controlling Chrome on remote machines via tunnels (e.g., traforo).