I open-sourced patchright-cli: a Patchright + real Chrome CLI for AI agents

I open-sourced a small project called patchright-cli The motivation was pretty simple: a lot of agents need to use a real browser, but the usual Playwright/CDP setup can get detected or blocked on some sites. I wanted a lightweight CLI that agents can use directly from a shell to drive real Google Chrome in a headful runtime, while using Patchright under the hood. The basic workflow is intentionally simple and both playwright-cli and patchright-cli are patched and interchangeable so agents don't need to "know" which one to use: 1. patchright-cli open google 2. patchright-cli snapshot 3. patchright-cli click e3 4. patchright-cli close It runs `playright-cli` under the hood but its patched so that it can run through patchright, and this does use a headful Google Chrome browser. It's much better at accessing many websites that agent-browser and many generic CDP driven tools get blocked on but is by no means perfect but works fine for my agents for simpler tasks. I did have codex add some simple hints to the agent to close tabs if they're resource intensive since otherwise agents default to simply leaving all tabs open. As well, it includes an associated skill for `patchright-cli` and `playwright-cli`. It’s fully free and open source. No hosted service, no API key, no paid tier. You can run it in a container or wire it directly into a Linux/Ubuntu-style sandbox. Would be curious what people here think, especially if you’re building agents that need browser access.