pagerunner, from Enreign, is an MCP server that connects AI agents to a live Google Chrome session to automate authenticated web tasks. The tool uses the Chrome DevTools Protocol to let models issue browser commands, navigate pages, and extract semantic page data for agent-driven workflows. It includes session checkpoints, PII redaction, site intelligence, and a macOS menu bar app. The target users are developers, AI researchers, and power users who need agent access to real browser sessions.

What tasks can you actually use it for?

The tool enables agent-driven navigation and data extraction inside a real browser. By exposing Chrome through the Chrome DevTools Protocol, it lets models perform precise interactions and pull structured page information. Typical outcomes include multi-step navigation, authenticated content retrieval using the existing Chrome profile, and extracting semantic page structure so agents can summarize or act on live site content.

How reliable are its browser interactions and data extraction?

Precision comes from using a real Chrome session rather than a simulated environment. The CDP integration executes native browser commands, which reduces mismatches between headless automation and user-visible pages. Site intelligence features provide structural metadata that improves model interpretation. Checkpoint and snapshot tools let users reproduce or roll back browser states, which supports repeatable workflows and debugging when agents take unexpected actions.

What inputs does it accept and what are the limits?

The server connects only to Google Chrome and requires a local Chrome installation. It attaches to an existing Chrome profile so sessions, cookies, and saved credentials remain under the browser's control; the tool does not store passwords independently. The developer built PII redaction to mask sensitive fields, but granting agent access to active sessions still requires user caution because live accounts may expose confidential content.

Is it practical to set up and integrate into developer workflows?

Installation and integration target technical users and developer environments. The tool offers Homebrew, Cargo, and pre-built binaries for macOS, Linux, and Windows, and a native macOS menu bar app for session monitoring. Registering it as an MCP server is done via configuration (for example, adding an entry to the agent's configuration file), making it suitable for researchers and power users who maintain local MCP setups and require browser-backed agent capabilities.

A practical choice for technical users who accept operational oversight

The tool is a practical option for developers and researchers who need agent-driven access to real web sessions; it improves reproducibility and in-browser control for scripted tasks. One limitation is the operational risk from granting agents live-session access, so restrict agent scope and audit outputs frequently. For technical teams comfortable with MCP configuration, the tool delivers usable browser-based automation when paired with disciplined verification.