Playwriter

Playwriter is a revolutionary developer tool that solves the fundamental problem of web automation for AI agents: access. Traditional methods force agents to operate in a sterile, isolated browser environment—a fresh Chrome instance with no logins, no extensions, and no cookies, which is instantly flagged by bot detection systems. Playwriter takes the opposite approach. It gives AI agents direct, programmable control over your actual, everyday Chrome browser session through a simple Chrome extension and CLI. This means your agents can interact with websites where you're already logged in, with your preferred extensions active, and with your existing cookies and local storage intact. It leverages the powerful Playwright automation API, exposing it through a single, flexible tool for AI clients via the Model Context Protocol (MCP). The result is robust, human-like browsing that bypasses common automation blocks, uses far less memory than spawning new browser instances, and enables deep debugging and inspection capabilities. It's open-source (MIT licensed), runs entirely locally, and is designed for developers and power users who need reliable, collaborative automation that works with the web as it exists for real users.

Your Real Browser Session

Playwriter's core feature is granting AI agents access to your live Chrome session. Instead of launching a separate, "headless" browser that lacks identity and history, the agent operates directly in your open browser tabs. All your active logins, saved cookies, browser extensions (like ad blockers or password managers), and cached data are immediately available. This eliminates the friction of re-authenticating for every task and dramatically reduces the chance of being detected as a bot, as the browser fingerprint matches your legitimate, daily use.

Full Playwright API via a Single Tool

Unlike other MCP implementations that expose a limited, predefined set of browser actions (like "click" or "type"), Playwriter provides agents with the entire Playwright automation library through one execute tool. This means the agent can write and run any valid Playwright code—from complex interactions and network interception to performance profiling. This approach avoids "schema bloat" from dozens of individual tool definitions, keeping context usage low for AI models and providing maximum flexibility for tackling any web task.

Advanced Debugging & Inspection Suite

Playwriter transforms your browser into a debuggable runtime for AI agents. It includes a full-featured debugger where you can set breakpoints to pause execution, live-edit code on the fly, and step through commands. The tool also provides network interception to monitor and modify HTTP requests/responses and built-in screen recording to capture video of the agent's actions. This suite is invaluable for developing, testing, and understanding complex automation workflows.

Lightweight Accessibility Snapshots

Instead of relying on large, inefficient screenshots (often 100KB+), Playwriter generates compact accessibility snapshots (5-20KB). These snapshots provide the AI agent with a structured, semantic view of the page, including element roles, names, states, and a logical reading order. This allows the agent to understand the page layout and interact with elements reliably using ARIA attributes, all while consuming minimal context tokens and speeding up processing.

Automated Testing & QA with Real User Data

Developers and QA engineers can use Playwriter to create and run automated tests that execute within a genuine user context. Since the agent uses your real browser session, tests can automatically navigate through authenticated areas of web applications, test features behind paywalls, or verify flows that depend on specific cookies or local storage—scenarios that are cumbersome or impossible to test with traditional, isolated automation frameworks.

AI-Powered Research & Data Extraction

For researchers, analysts, or anyone needing to gather information from multiple web sources, Playwriter enables AI agents to conduct deep, session-aware research. The agent can log into academic journals, navigate complex multi-page search results, accept cookie consents, and extract structured data—all while maintaining login sessions across different websites. You collaborate by handling unexpected pop-ups or CAPTCHAs in real-time.

Repetitive Web Task Automation

Automate tedious, daily web tasks without writing complex scripts. An AI agent powered by Playwriter can log into your admin dashboard, generate and download reports, fill out recurring forms, or monitor specific pages for changes. Because it works in your browser, it can access internal company tools or portals that require specific corporate authentication, turning multi-step manual processes into a single command.

Collaborative Web Development & Debugging

Front-end developers can work alongside an AI agent to debug issues. You can instruct the agent to reproduce a bug, intercept network calls to inspect API payloads, set breakpoints on specific user interactions, and take snapshots of problematic states. This turns the AI into a pair-programming partner that can manipulate the browser and inspect its state at a level of detail that goes beyond simple chat instructions.

How does Playwriter handle security and privacy?

Playwriter is designed with a strong local-first principle. All connections are made via a WebSocket relay running on your local machine (localhost:19988). No browser data, automation commands, or screenshots are sent to any remote server. The extension only communicates with the local relay, and you have full control over which tabs the agent can access by clicking the extension icon to attach or detach. You are always in the driver's seat.

Can I use Playwriter with any AI assistant or IDE?

Yes, Playwriter is built on the open Model Context Protocol (MCP), which is supported by a growing number of clients. This includes popular AI-powered IDEs like Cursor and Windsurf, AI assistants like Claude Desktop, and code editors like VS Code with appropriate extensions. As long as your client supports MCP, you can connect it to the Playwriter server and begin automating.

What happens if the agent gets stuck or encounters a CAPTCHA?

This is where Playwriter's collaborative nature shines. You are watching the browser in real-time. If the agent encounters a CAPTCHA, a unexpected pop-up, or simply gets stuck in a loop, you can manually intervene. You can solve the CAPTCHA yourself, click the necessary button, or even temporarily detach the extension from that tab to fix the state manually. Once you re-enable the extension, the agent can continue from the new state.

Is Playwriter only for Chrome?

Currently, Playwriter is implemented as a Chrome extension, so it requires the Chrome or a Chromium-based browser (like Brave, Edge, or Arc) to function. The automation API it exposes is Playwright, which is cross-browser, but the mechanism for attaching to your live session is specific to the Chrome DevTools Protocol (CDP) via the browser's extension system.