PraisonAI Browser Agent

Side panel browser automation. Navigate, click, type, run JavaScript, extract page data, and capture screenshots.

PraisonAI Browser Agent — a side panel toolkit for driving web pages by hand. Open the side panel and control the active tab directly: go to a URL, click an element by CSS selector, type into a field, run a JavaScript expression, read the page's console output, pull its structure out as JSON, or capture a screenshot. Everything runs locally in your browser using the Chrome DevTools Protocol. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ WHAT YOU CAN DO ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ • Navigate — enter a URL and load it in the current tab • Click — target any element with a CSS selector • Type — enter text into an input or textarea by selector • Scroll — move the page up or down • Run JavaScript — evaluate an expression in the page and see the result • Extract Data — get the page title, URL, headings, links and images as JSON • Console Logs — read console output and uncaught errors from the page • Screenshot — capture the page from the panel, a keyboard shortcut, or the right-click menu, then save the image from the panel • End Session — detach from the tab and clear Chrome's debugging banner • History — the panel remembers the actions you ran, stored on your device ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ HOW TO USE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1. Click the PraisonAI icon in your toolbar to open the side panel (or press Ctrl+Shift+P, Cmd+Shift+P on Mac, or right-click and choose "Open PraisonAI panel") 2. Fill in a URL, selector, text or expression and press the button 3. Results appear in the Output box below 4. Press "End Session" when you are finished with a tab The panel belongs to the window it was opened in and drives that window's active tab. While an action is running the buttons are disabled, so two actions never overlap on the same tab. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ KEYBOARD SHORTCUTS ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ • Ctrl+Shift+P (Cmd+Shift+P) — open the side panel • Ctrl+Shift+S (Cmd+Shift+S) — capture a screenshot ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ GOOD TO KNOW ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ • Automation uses the Chrome DevTools Protocol, so Chrome shows a "PraisonAI Browser Agent started debugging this browser" banner while the panel is driving a tab. Press "End Session", or close the banner, to stop. • Chrome does not allow automation on chrome:// pages, the Chrome Web Store, or other extensions' pages. The panel tells you when the current tab is off limits. On a brand-new, empty tab it asks you to open a website first. • Console output is captured from the moment the panel connects to a tab. Reload the page to capture its start-up logs. "End Session" or closing the tab clears what was captured. • Screenshots are shown in the panel with a Save link. Nothing is written to your disk unless you save it yourself. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ WHAT THE PERMISSION WARNING MEANS ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Chrome shows two warnings at install: "Access the page debugger backend" and "Read and change all your data on all websites". Both come from the debugger permission, which is what the Chrome DevTools Protocol requires. It is the honest ceiling of what the permission allows, and we would rather explain it than let it surprise you. What the extension actually does with it: • It attaches to one tab at a time, and only in response to something you did: a button in the panel, the screenshot shortcut, or the right-click menu • Chrome shows a "started debugging this browser" banner the whole time • Press "End Session", or close the panel, and the session ends immediately. A screenshot taken by shortcut or right-click with the panel closed detaches as soon as it is done, so it leaves no session behind • It declares no host permissions and installs no content script, so it has no standing access to any site and does nothing in the background • It makes no network requests, so nothing it reads can go anywhere • Actions report what actually happened. A URL that does not resolve, a selector that matches nothing, and text a field would not accept are all reported as failures — never as success. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ PRIVACY ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ • No account and no sign-in • No servers — the extension makes no network requests of any kind • No analytics and no tracking • Three things are stored, all locally, using Chrome storage: your action history, the most recent screenshot, and the console output captured from the tab you are working on. Nothing is transmitted. • No host permissions are declared. Chrome still shows "Read and change all your data on all websites" at install, because the debugger permission implies it — see "What the permission warning means" below. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ SUPPORT ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ • Documentation: https://chrome.praison.ai • Issues: https://github.com/MervinPraison/praisonai-chrome-extension/issues