Is it safe to let an AI control my browser?
It is designed so you stay in charge. Nothing runs until you switch it on, a one-click pause stops everything, risky actions (payments, deleting, sending, passwords, opening new sites, file uploads, running JavaScript) open an approval window that a web page cannot see or click, and organisation admins can block sites and disable tools. No system can promise to stop every prompt-injection attack, so PagePilot shows warnings and asks before anything irreversible.
Which Claude apps does it work with?
Any MCP client that supports remote servers with OAuth sign-in, including Claude Code (claude mcp add --transport http) and Claude Desktop or claude.ai custom connectors where available on your plan. You add one URL, sign in once in your browser, and Claude gets the browser tools.
What can the service see?
To let Claude use your browser, page content passes through our relay in transit to your Claude app. It is not stored. What we do store is account data and an audit log of metadata only (which tool ran, on which site, when), never page content, typed text or files. Teams that need full control can self-host the relay.
Do I need to install anything besides the extension?
No. In cloud mode there is no local helper: you add the browser extension, sign in by approving a short pairing code, and add the connector URL in Claude.
Which browsers are supported?
Google Chrome and other Chromium-based browsers such as Microsoft Edge and Brave.
Can my company restrict what employees can do?
Yes. Owners and admins set a policy for the whole organisation: sites that are blocked or the only ones allowed, tools that are switched off, a minimum approval level, and whether advanced tools are allowed. Policy is enforced both by the service and inside the extension, and every tool call is recorded in the audit log.
What does it cost?
PagePilot is free during early access. Team features will be priced once the product leaves early access; we will tell you well before anything changes.