What problem does it solve? When multiple AI agents share a single Playwright MCP connection, every browser call hits the same browser instance, causing state contamination and blocking. This Skill installs and configures a pooling MCP proxy that gives each agent its own isolated browser session. ## Core Features & Use Cases - Session-based isolation: Agents acquire a session_id via browser_pool_acquire, pass it to every browser_* call, and release it with browser_pool_release when done. - Self-sizing pool: Keeps one warm standby browser for instant acquisition, grows on demand, and supports an optional maxConcurrent hard cap with queued acquires and configurable timeouts. - Guided setup wizard: Detects Cursor vs Claude Code, verifies Node.js and @playwright/mcp prerequisites, installs browser binaries, deploys the server, and registers it as an MCP server. - Use Case: Three agents scraping different sites in parallel each acquire their own session, navigate and snapshot independently, then release their browsers without interfering with one another. ## Quick Start Ask your agent to set up the playwright pool and follow the prompts to choose a browser, timeout, and concurrency cap.
