p agent extension

Local browser control bridge for AI agents via Native Messaging

p agent extension is a companion browser extension for compatible AI agents running through a local MCP or Pi bridge. It provides a visible, locally controlled interface that lets a configured AI agent work with browser tabs and page controls. Features: • Open, select, group, manage, and close browser tabs • Inspect visible page structure and interactive elements • Click buttons, enter text, and select page options • Capture screenshots and page information • Run supported browser automation commands • Show active automation through an on-page HUD, element badges, cursor indicators, and a visible tab group How it works: A separately installed local bridge connects the extension to compatible AI-agent software through Native Messaging or a localhost connection. The local bridge must be running for agent-control features to work. Transparency and privacy: When instructed by the connected agent, the extension can read tab titles and URLs, inspect page content and structure, capture screenshots, and interact with page controls. The extension displays controlled activity on the page so users can see which tab and elements the agent is working with. The extension does not request cookie access, browsing-history access, or persistent browser-storage access. It contains no advertising, affiliate tracking, or cookie-management functionality. Information handled by the connected AI agent is subject to that agent’s configuration and privacy practices. Only use the extension on pages you are authorized to automate.