Web Draw by Olib AI

Let an AI agent read and drive the page you are looking at, as text, without screenshots.

Web Draw lets an AI assistant read and operate the browser tab you already have open, using text instead of screenshots. The extension reads the visible part of the page and turns it into a compact text view: every button, link, field and menu item, each with a short handle the assistant can act on, plus the state that matters. Which field is required. Which one is invalid. What is disabled. What a dialog says. What the page replied after a click. Because it works on the browser you are already signed in to, it uses your existing sessions. There is no separate login, no headless browser, and no copying of credentials. HOW IT WORKS Web Draw has two parts. This extension reads the page. A small local server connects it to your AI client over the Model Context Protocol. After installing, add the server to your MCP client configuration: "web-draw": { "command": "npx", "args": ["-y", "@olib-ai/web-draw-mcp"] } The toolbar badge shows ON once the two are connected. You need Node.js 20 or newer. WHAT IT IS GOOD AT - Filling and submitting real forms, including long multi step ones - Reading dense pages such as search results, dashboards and data tables - Handling dialogs, cookie walls, overflow menus and controls inside iframes - Telling the assistant when a page refused a submit, and which fields it named, so the next attempt fixes the right thing A typical page costs a few hundred tokens to read this way, rather than the several thousand a screenshot needs, and the assistant gets something it can click rather than a picture it has to guess at. PRIVACY Everything runs on your machine. The extension talks only to 127.0.0.1 on the port you configure. Page content goes to your AI client through your own local server. Nothing is sent to Olib AI, and there is no analytics or tracking in this extension. WHY THE PERMISSIONS - Access to all sites and tabs: to read the page you point it at and act on it. It reads a page only when your assistant asks it to. - Scripting: to load the reader into a frame that does not have it yet, so pages open before install still work. - Debugger: used only as a fallback, when a page ignores ordinary synthetic events. It drives keyboard and mouse input locally. Your browser shows its standard notification while this is active. - Web navigation: to find the frames inside a page. - Storage and alarms: to keep your connection settings and hold the local connection open. NOT WHAT YOU NEED? Web Draw is for driving one browser you can see, with your own logins. If you want large scale headless automation, many parallel browser contexts, or antidetect browsing, use Owl Browser instead: https://www.owlbrowser.net