SKI
On-device voice interface that lets developers speak to coding agents like Claude Code and Codex and hear them talk back.
Community:
Product Overview
What is SKI?
SKI turns typed prompts into spoken conversation for coding agents. Instead of typing instructions and scrolling through output, developers talk to Claude Code, Cursor, Codex, Gemini CLI, Windsurf, or OpenClaw out loud and hear the agent's reply in a synthesized voice. The entire loop — speech recognition and voice output — runs locally on the user's Mac or PC, so nothing spoken is ever transmitted to a server. Beyond coding, SKI doubles as an offline meeting recorder and, through its paid AgentCall feature, can send the agent into a live video call as a speaking participant or silent notetaker.
Key Features
Fully On-Device Voice Loop
Speech-to-text and neural voice synthesis both execute on the local machine, so no audio ever leaves the device and the app keeps working offline.
Full-Duplex Barge-In
Echo cancellation lets users talk over the agent mid-reply and still be heard clearly, removing the need for push-to-talk controls.
Cross-Agent, Multi-Project Routing
One integration connects to several coding agents at once, with each bound project speaking in its own distinct voice for easy switching.
Live Status and Approve-Before-Send
A real-time indicator confirms the agent is actively listening, while an optional review mode holds transcripts in an editable bubble until the user confirms sending.
On-Demand Screenshot Capture
A hotkey attaches a screen grab to the next spoken request, and the agent can also request its own screenshot when it needs visual context.
Offline Meeting Recorder with Calendar Auto-Join
Captures mic and system audio locally with unlimited length and no bot joining the call, and can automatically dispatch the agent to calendar-linked meetings.
Use Cases
- Hands-Free Coding Sessions : Developers issue coding instructions verbally and hear results narrated back instead of reading through terminal output.
- Multi-Repo Task Switching : Engineers managing several codebases address the right project by voice, using each project's distinct voice as an audio cue.
- Local Meeting Transcription : Teams record stand-ups, reviews, or client calls entirely on-device and export the transcript without relying on a cloud transcription service.
- Visual Context Sharing : Developers pair a spoken request with an instant screenshot so the agent can see exactly what's on screen before acting.
- Voice-First Accessibility : Developers who prefer speaking over typing, or who have limited typing capacity, can build and debug software conversationally.
- AI Agent Meeting Participation : Through AgentCall, the coding agent joins a Zoom, Meet, or Teams call to speak, present, or quietly take notes on the team's behalf.
FAQs
SKI Alternatives
ElevenLabs
Advanced AI-driven platform specializing in lifelike text-to-speech, speech-to-text, voice cloning, and conversational voice agents across multiple languages.
Cursor
AI-powered code editor built on VS Code that accelerates software development with intelligent code generation, refactoring, and contextual codebase understanding.
Truecaller
A leading global caller ID and spam blocking app that uses AI and community data to identify calls, block spam, and enhance communication security.
Xiaomi MiMo
Xiaomi's full-stack agent model suite covering frontier reasoning, omnimodal perception, and expressive speech synthesis — built for the agentic era.
Sesame AI
Advanced AI voice model delivering natural, expressive, and context-aware conversational speech synthesis.
Deepgram
A leading voice AI platform that provides speech-to-text, text-to-speech, and speech-to-speech capabilities for developers.
SoundHound AI
Advanced voice AI platform delivering highly accurate, customizable conversational experiences with integrated generative AI and music recognition.
Cartesia AI
The fastest ultra-realistic voice AI platform enabling real-time voice synthesis, cloning, and infilling with high fidelity and low latency.