AgentsRoom.dev : AI Dev Remote Control (Chrome companion)

The official Chrome companion for AgentsRoom.dev, the multi-agent command center for AI coding CLIs. /!\ You must download AgentsRoom desktop app (Windows, MAC or Linux) /!\ 1. Send pages and elements to your agents Browse any website. Open the extension, and send the full page, a specific element or a selected block of text straight into your active agent's conversation in the AgentsRoom desktop app. Your agent receives the content with the page URL and context already included, ready to act. No copy-paste, no tab switching, no re-explaining. "Point on page" is one click, or one shortcut (Alt+Shift+P by default): hover an element, click it, and it lands in your agent's message box with its HTML and a selector. "Send page to my desktop" drops the whole page over your local loopback, so the HTML never travels through a server. 2. File bugs and requests in one click Hover any element on the page, click to pin it, type what is wrong, and hit send. A structured ticket lands in your AgentsRoom backlog with the pinned element, the page URL and an optional screenshot already attached. Titles are generated automatically so the ticket is immediately actionable. Two modes: - Logged-in mode: tickets go directly into one of your projects. - Public backlog mode: share a backlog link with anyone, no account required. Clients and testers click the extension, point at what is broken, and the ticket arrives in your backlog. For agencies, this replaces vague feedback emails with precise, visual reports you can assign to agents straight away. 3. Auto-switch projects by URL Associate a URL pattern with a project in AgentsRoom. From that point, navigating to a matching page in Chrome automatically switches the desktop app to that project. Working across multiple client websites? Each time you change tabs, AgentsRoom follows. No more hunting through the sidebar to find the right project. What makes it different: One Chrome per account. Chrome runs one copy of the extension per profile, so each profile becomes its own named target. One agent works on your pro account while another works on your personal one, at the same time, and neither sees the other's session. Your browser, on any machine. The extension talks to the AgentsRoom relay, the same one the mobile app uses. An agent on the Mac at the office can drive the Chrome on the PC at home. Every message and every screenshot is end-to-end encrypted with a key only your devices hold. Every agent CLI. It is part of the AgentsRoom Browser MCP server, so Claude Code, Codex, and more : including the ones that ship no browser extension of their own, like OpenCode or Grok Build. You always know. Driving is off by default. The first action opens a fresh tab, never the one you are reading. Chrome shows its own "AgentsRoom started debugging this browser" banner for the whole session, and switching the toggle off closes every session and detaches the debugger. The embedded browser of the desktop app stays the default target for testing your localhost. This Chrome is an extra target, chosen per agent, for the work that needs your real accounts. Who it is for - Solo builders: file issues on your own staging site and let your AI agents fix them while you keep browsing. Let an agent check a deploy behind your SSO in the Chrome where you are already signed in. - Agencies: share a public backlog link with your client. They report issues by pointing directly on the page. You get structured, visual tickets instead of scattered emails. Name a Chrome profile per client and point the right agent at it. - Teams: product managers, designers and QA testers can file tickets from any environment, with no setup beyond installing the extension. A QA agent can verify the staging site in a profile signed in as the test user. Privacy The extension requests only what it needs: - activeTab, tabs: reads the current page URL and your selected text when you open the popup, and finds the tab an agent is driving. - scripting: injects the element picker on the tab you interact with. - storage: saves your session, your preferences and the name you gave this browser, locally. - identity: signs you in to your AgentsRoom account. - alarms: keeps the connection to the relay alive while driving is on. - debugger : requested only when you switch on "Let my agents drive this browser", and only attached to a tab while an agent is actually driving. Never at install time. Host permission for agentsroom.dev and your public backlog domains: to create tickets and authenticate. Optional host permission for localhost and 127.0.0.1: only if you use "Send page to my desktop", which talks to the desktop app on your own machine. Driving is off by default and switches off with one click. What an agent sends to a driven Chrome travels through the AgentsRoom relay as encrypted blobs: the relay cannot read a URL, a page or a keystroke. No background tracking. No analytics. No third-party data sharing. About AgentsRoom AgentsRoom is a visual command center for running AI coding agents in parallel on your codebase. It includes a visual backlog, 14 specialized agent roles, real-time terminals, an embedded browser and a Browser MCP server, an end-to-end encrypted relay, a prompt library and companion iOS and Android apps. Download the desktop app at https://agentsroom.dev. Support: support@agentsroom.dev