easi.run gives AI agents real Chrome sessions on your own Linux x86-64 server through MCP or REST. Developers can watch runs live and take control for sign-ins, two-factor authentication and approvals, then return control to the connected agent. The agent provides the reasoning; easi.run provides the browser environment.
Features include encrypted saved profiles, isolated sessions, file transfers, action recordings and audio. It is free software under AGPL-3.0-only and requires rootful Docker. You supply the server and your own agent/model; infrastructure and model costs are separate. It is useful for browser workflows that need both agent actions and human participation, including authenticated websites and downloading finished files.
Documentation: https://docs.easi.run
Source: https://github.com/julianosig/easi.run
Live demo: https://easi.run/demo
Submitted by creator Juliano Gouveia with AI assistance, checked against official product information.