AAEP: Agent Accessibility Event Protocol

AAEP is an open protocol that lets AI agents tell assistive technology what they are doing, in a format any screen reader, voice control, or alternative input device can understand. For the more than one billion people who use assistive technology, AI agents are often unusable: state changes are invisible, tool calls are silent, streaming output floods the screen reader, and irreversible actions happen without a surfaced confirmation. Every team solves this differently or not at all, because there is no shared language between agents and assistive technology. AAEP is that shared language. It defines twelve standard event types, three conformance levels, and an open test suite that verifies any claim of support. The spec is CC-BY-4.0, the code is MIT, and no single vendor can take it private. v1.0.0 shipped in June 2026 with a five-year support window. Reference packages are on PyPI and npm, and it works with Microsoft Agent Framework, LangChain, Semantic Kernel, the Anthropic SDK, OpenAI, and custom agents, across Narrator, NVDA, JAWS, VoiceOver, and TalkBack.