Chirply Browser Agent Beta

Turn the page you are already viewing into a user-directed browser workflow. Chirply Browser Agent inspects the active page, prepares a visible action plan, and runs it only after you approve it. CHIRPLY ACCOUNT REQUIRED Requires an active Chirply account. This extension does not work as a standalone product. HOW IT WORKS Open the Browser Agent side panel, describe what you want accomplished, and preview the task. Browser Agent inspects the page you deliberately selected, including visible text and safe labels for links, fields, menus, and buttons. Before anything changes, you can review: • The exact page title and URL • Every proposed browser action • Text that will be entered into fields • Existing non-secret form values • Any click, navigation, submission, or outward action WHAT BROWSER AGENT CAN DO With an approved plan, Browser Agent can: • Extract visible page content, links, tables, listings, and selected information • Turn webpage information into structured local results • Fill text fields and select available options • Click visible controls after approval • Scroll and wait for page updates • Open an approved HTTPS destination • Assist with compatible workflows on websites where you are already signed in • Help prepare and complete compatible social-media workflows through the page itself Browser Agent works through the webpage and Chrome session you already control. This can help with research, data extraction, repetitive data entry, internal tools, and compatible browser workflows that do not have a dedicated API integration. PRIVATE, LOCAL PLANNING On supported Chrome 138+ devices, Browser Agent can use Chrome’s built-in on-device language model to turn your instruction into a typed action plan. If the on-device model is unavailable, a deterministic planner packaged with the extension handles supported commands instead. Your task instruction, page snapshot, form values, and extracted results stay inside the extension. They are not sent to Chirply or to a third-party AI provider. APPROVAL AND CONTROL Every run begins with an instruction you provide on the active page. A plan cannot approve itself. Every click receives the strongest outward-action warning. The label shown for the control is rebuilt from the inspected page rather than trusted from generated model text. Approval is bound to the inspected page using a SHA-256 state token. If the page changes unexpectedly after approval, Browser Agent stops the run and requires a fresh inspection instead of guessing through the changed page. You can stop an active run from the side panel at any time. NARROW BROWSER PERMISSIONS Browser Agent uses temporary active-page access. It does not request permanent access to every website. The extension does not request access to: • Your general browser history • Website cookies • Saved passwords • Clipboard contents • File-upload values • Every website you visit • Remotely hosted executable code Browser Agent does not export website sessions, bypass access controls, solve CAPTCHAs, defeat paywalls, or run arbitrary JavaScript supplied by a remote server. PAIRING AND AUDIT HISTORY Pairing uses a single-use code generated inside your Chirply account. Each paired browser receives a separate revocable device credential. Chirply receives only a redacted run record containing: • The page origin without its path or query • A coarse website label • Packaged action types and risk levels • Approval and completion status • Result count • A generic error code when applicable Chirply does not receive the task instruction, full page URL, page snapshot, extracted content, form values, website passwords, cookies, or full results. Workspace owners can review the redacted run history and revoke any paired browser from Chirply. Browser Agent is a launch beta. Website compatibility varies. You remain responsible for ensuring that you are authorized to access the content and perform the workflows you choose, including complying with each website’s rules and terms.