MCP server for AI-agent control of screen readers — drive gestures, capture speech/braille, and automate add-on testing. NVDA today; JAWS/TalkBack bridges planned.
Or install with pip: pip install screen-readers-mcp
README Excerpt
An MCP server that lets an AI agent **drive NVDA**: send keyboard gestures, read back what NVDA speaks (and brailles), and introspect its state. The first use case is **functional testing of NVDA add-ons** — replacing manual testing — but the same primitives support a wider range of agent-driven NVDA workflows.