Playwriter

Stop fighting with headless browsers and bot detection. Playwriter is the game-changing bridge that lets your AI agents operate directly within your real, logged-in Chrome browser. It's a Chrome extension and CLI that gives any MCP client (like Cursor, Claude Desktop, or VS Code) the full power of the Playwright automation API, but executed in your existing browser session. This means your agents have immediate access to all your extensions, cookies, and logins—no fresh Chrome instance, no extra memory load, and no instant red flags from websites. It turns the brittle, limited browsing of today's AI into a seamless, powerful, and collaborative experience. Built for developers, researchers, and power users who need their AI assistants to interact with the modern web as a human would, Playwriter is open-source (MIT licensed) and runs entirely locally on your machine, putting you in full control.

Your Real Browser Session

Playwriter doesn't spawn a new, sterile browser. It attaches directly to your existing Chrome tabs via a debugger connection. This means every website you're already logged into—Gmail, GitHub, admin dashboards—is instantly available to your AI agent. Your extensions (like password managers or ad blockers) are active, and your cookies are present, completely bypassing the bot detection that plagues fresh, headless instances. It's browsing with your identity, without the overhead.

Full Playwright API Through a Single Tool

Instead of limiting agents to a fixed set of pre-defined tools (like "click" or "type"), Playwriter exposes one powerful execute tool. Your agent can run any valid Playwright code, unlocking the entire automation library: complex selectors, network interception, performance profiling, and setting breakpoints. This keeps the AI's context window lean by avoiding schema bloat from dozens of tool definitions and grants unparalleled flexibility.

Advanced Debugging & Collaboration Suite

Playwriter is built for real-world debugging and human-AI teamwork. It includes a live debugger with breakpoints, live code editing, and network request interception. You can watch the agent work in real-time on your screen. When it hits a CAPTCHA or a consent wall, you simply solve it and let the agent continue. If it gets stuck, you can temporarily disable control, fix the page manually, and re-enable it—seamless collaboration.

Lightweight Accessibility Snapshots & Recording

Forget massive, context-consuming screenshots. Playwriter generates compact accessibility snapshots (only 5-20KB) that provide the AI with a structured, semantic view of the page. It also includes visual labeling for elements and can record full video of the agent's session. Combined with network interception, this gives your AI deep, efficient insight into page state and behavior.

Automated Web Testing & Debugging

Developers can instruct their AI assistant to run complex, multi-step user flows on their already-logged-in development or staging environments. The agent can click through UIs, fill forms, intercept API calls to verify payloads, and profile performance—all within the exact browser context a real user would have, making tests more reliable and context-aware.

AI-Powered Research & Data Extraction

Researchers and analysts can task an AI with gathering information from websites that require login or have complex interactive elements. Since Playwriter uses the real browser, it can navigate paywalls, logged-in databases, and JavaScript-heavy dashboards that are completely inaccessible to traditional, headless agent tools.

Routine Task Automation

Automate repetitive daily web tasks without writing a single line of code yourself. Have your AI agent log in to platforms, generate reports, schedule posts, or monitor status pages. Because it works in your browser session, it can handle two-factor authentication flows where you approve the login prompt.

Enhanced AI Assistant Capabilities

Supercharge coding assistants like Cursor or Claude. Instead of just writing hypothetical code, your AI can now execute Playwright scripts to interact with live documentation, pull examples from a web API, test a component in Storybook, or even debug a live web application by manipulating the DOM and inspecting network traffic in real-time.

Is my browsing data sent to a remote server?

No. Playwriter is designed with privacy and security first. All communication happens via a WebSocket relay running on localhost:19988 on your own machine. The Chrome extension, CLI, and your AI client communicate directly with this local relay. No browsing data, credentials, or session cookies are ever transmitted to any remote server.

How does Playwriter avoid bot detection?

It avoids detection by not acting like a typical bot. It controls your genuine, user-initiated Chrome browser session, which already has a normal history, cookies, and likely some browser extensions installed. This fingerprint matches that of a regular human user, unlike a fresh, headless Chrome instance which lacks these signals and is easily flagged.

Can I use it with any AI or just specific MCP clients?

Playwriter works with any client that supports the Model Context Protocol (MCP), which is becoming a standard for AI tool integration. This includes popular environments like Cursor, Claude Desktop, VS Code with extensions, and more. The provided skill teaches these clients how to use Playwriter effectively.

What happens if the AI gets stuck or makes a mistake?

This is where the collaboration features shine. You watch the actions happen live in your browser. If the agent enters an error state or gets stuck in a loop, you can immediately click the extension icon to detach it from that tab (turning it gray), manually correct the situation, and then re-attach control. The agent can then continue from the corrected state.