MCP // SPEAK
MCP // SPEAK is a macOS MCP server that gives AI agents a voice — real-time speech via OmniVoice and native 'say' so agents announce status and ask questions out loud.
Last verified:
What is MCP // SPEAK?
MCP // SPEAK is a Model Context Protocol server for macOS that adds real-time voice synthesis to AI agents using OmniVoice and native macOS say. It enables agents to proactively announce status updates, ask clarifying questions, and deliver vocal feedback during pair programming sessions with sub-second latency.
MCP // SPEAK pricing
Pricing model: Freemium
MCP // SPEAK pros
- 10 distinct personality personas that can be customized and auditioned before deployment
- Sub-second latency synthesis on Apple Silicon with MPS acceleration
- Integrates with multiple AI clients (Claude Desktop/CLI, Cursor IDE, Google Antigravity, Windsurf, Codex)
- Interactive setup wizard and prompt compiler for easy configuration and system instruction generation
MCP // SPEAK cons
- macOS only—limited to Apple Silicon with MPS support
- Specialized use case (voice synthesis for AI agents) may have limited applicability beyond pair programming
- No obvious support for remote deployment or multi-user/collaborative scenarios
Frequently asked questions about MCP // SPEAK
What TTS engines does MCP // SPEAK support?
It uses OmniVoice and native macOS say for speech synthesis.
How do I install MCP // SPEAK?
Clone the repository with `git clone https://github.com/fellowgeek/mcp-speak.git`, then run `python3 setup.py` to launch the interactive setup wizard.
What AI clients are supported?
Google Antigravity, Claude CLI, Claude Desktop, Cursor IDE, Windsurf, and Codex Desktop/CLI.
Can I customize the voice personality?
Yes—10 pre-built personas are available (Sarcastic Senior, Eager Intern, Existential Emo, Pun Master, Tech Priest, Agent Smith, Gothic Poet, Nature Narrator, Head Chef, Neutral Mainframe), and you can generate custom system instructions via the prompt compiler.