XAgent has two products. Desktop organizes files, tasks, and agents around a Project. Browser awakens analysis and action only on pages matched to an authorized Project. Each works on its own, and they hand work off through explicit permissions.
Desktop is the control plane for projects and agents. Browser is the action surface for real web pages. They share one clear Project boundary without exposing ordinary browsing, the file system, or private data to AI by default.
More than an enlarged chat box. Every Project keeps its own file context, task chain, execution agents, and permission history—from planning and tool use through delivery.
A complete, independent multi-tab browser first. Page and viewport analysis wakes only when the page matches an authorized Project, with preview, confirmation, and apply before web actions take effect.
Desktop defines the Project and task. Browser continues that work on authorized pages. Capability follows the Project boundary—not the entire device or every page you visit.
Files, tasks, agents, and resource grants belong to the active Project, so work can continue without restarting from zero.
AI tools appear only on explicitly authorized pages matched to the Project. Ordinary pages remain ordinary browsing.
Important changes are previewed before they apply, while grants, device bindings, and outcomes remain reviewable and reversible.
We detect your operating system and chip architecture automatically. Desktop and Browser use independent installers, so you can choose either or both.
xagent Desktop
xagent Browser
