SKI

Voice coding agent that listens and responds aloud on-device

Coding & Development Coding AssistantsFree

Visit website ↗

FreeStarting price

1.9KMonthly visits

5Popularity

About SKI

SKI enables developers to control coding agents entirely by voice. The tool installs locally and integrates with popular agents such as Claude Code, Cursor, Codex, and others. Users speak commands, and the agent processes them while responding verbally in real time. All speech recognition and voice synthesis occur on the user’s machine, ensuring no audio or transcripts leave the device. The interface appears as a minimal floating pill that expands when active and retracts when idle. It supports full-duplex communication, allowing users to interrupt the agent mid-response without push-to-talk. Projects can be bound to individual voices, and the tool includes features like screen capture, meeting transcription, and calendar integration. Meetings can be recorded locally with speaker-diarized transcripts, and agents can join calls as participants or silent notetakers.

Key features

On-device speech-to-text and neural voice synthesis
Full-duplex echo cancellation and barge-in
Multi-project voice binding and routing
Screen capture integration with agent commands
Live agent status indicator
Calendar auto-join for meetings
Offline meeting recorder with unlimited retention
Global hotkeys and draggable live-glass pill

Use cases

Conducting hands-free coding sessions with voice commands
Recording and transcribing local meetings with speaker tags
Joining meetings as an agent participant or notetaker

Pros

Runs entirely on-device with no cloud dependency
Supports multiple coding agents with one installation
Full-duplex voice interaction without push-to-talk
Local meeting transcription and recording
Customizable hotkeys and minimal UI footprint

Cons

No cloud processing or remote access
Requires local installation per operating system
Limited to supported coding agents

Frequently asked questions about SKI

What is SKI and what does it do?

SKI enables developers to control coding agents entirely by voice. It allows users to speak commands to supported agents such as Claude Code, Cursor, or Codex, which then process the requests and respond verbally in real time. All speech recognition and voice synthesis occur locally on the user’s machine, ensuring privacy.

Who is SKI designed for?

SKI is designed for developers who want to streamline their workflow by using voice commands to interact with coding agents. It is particularly useful for those who prefer hands-free coding or need to multitask while working on projects.

Does SKI require an internet connection to work?

No, SKI operates entirely offline. Both speech-to-text and voice synthesis run locally on the user’s machine, and no audio or transcripts are sent to the cloud. This ensures that all data remains private and secure on the device.

Which coding agents does SKI support?

SKI supports multiple coding agents, including Claude Code, Cursor, Codex, Gemini CLI, OpenClaw, and Windsurf. Users can connect their preferred agent with a single click during setup.

Can SKI be used for meetings and recordings?

Yes, SKI includes features for meeting transcription and recording. It can capture local audio from meetings on platforms like Zoom, Meet, or Teams, transcribe them on-device with speaker diarization, and even allow agents to join meetings as participants or silent notetakers.

How do I get started with SKI?

Getting started with SKI involves downloading and installing the tool from the official website. The setup process includes configuring your microphone, selecting a voice, and setting a global hotkey. Once installed, you can connect your preferred coding agent and begin using voice commands immediately.