Mcp Listen
Add microphone capture and speech-to-text so your agent can react to live voice input on the dev machine.
Overview
mcp-listen is an MCP server for the Build phase that captures microphone audio and returns speech-to-text so your AI agent can listen.
What is this MCP server?
- Microphone capture exposed to MCP clients
- Speech-to-text pipeline for agent-consumable text
- stdio npm package mcp-listen v0.1.1
- Targets local dev agents that need ears, not batch transcription SaaS
- Lightweight companion to vision or chat-only MCP stacks
- npm package mcp-listen version 0.1.1 with stdio transport
Community signal: 5 GitHub stars.
What problem does it solve?
Text-only agents cannot hear standups, pair-programming chatter, or your own voice commands without you typing transcripts by hand.
Who is it for?
Solo builders prototyping voice-driven dev assistants or accessibility flows on a local machine with MCP hosts.
Skip if: Teams needing compliant call-center transcription, phone integrations, or production-scale streaming STT without reviewing v0.1.x limitations.
What do I get? / Deliverables
After install, your agent can pull live or recorded mic input through MCP and work from transcribed speech in the same session.
- Mic capture and STT tools available to your agent
- stdio MCP registration for voice-driven workflows
- Transcribed text usable in downstream agent tool calls
Recommended MCP Servers
Journey fit
Voice I/O is built when you extend the agent product surface—canonical placement is build while wiring multimodal tooling. Agent-tooling fits because listen/STT is infrastructure for your assistant, not a go-to-market or analytics workflow by itself.
How it compares
Local mic + STT MCP utility, not a full voice SaaS or telephony integration.
Common Questions / FAQ
Who is mcp-listen for?
Developers using Claude Code, Cursor, or generic MCP clients who want microphone and speech-to-text tools wired into their agent on a dev machine.
When should I use mcp-listen?
Use it while building voice-augmented agent demos, hands-free commands, or accessibility experiments—not as your primary production transcription service.
How do I add mcp-listen to my agent?
Install npm package mcp-listen, grant OS microphone permission, add the stdio server entry in your MCP config, restart the host, and invoke the listen/STT tools from chat.