Skip to Content

Autonomy

Autonomy is the accessibility layer for agentic computer use: a local MCP/CLI runtime that lets an AI agent see, act, speak, and browse through your Mac’s real accessibility setup instead of guessing from pixels.

It’s built first for blind and low-vision users working screen-off — no monitor to watch, no chat window to find — who want an agent that narrates what it’s doing, asks before anything risky, and works through their actual assistive technology: VoiceOver, keyboard-first control, Switch Control, and more. Agent builders, agent hosts, and operators get the same runtime underneath: one MCP server, one CLI, and one consent model enforced identically for every connected agent.

Last updated on