Cue — ИИ-копилот поверх экрана для встреч

★ 1.4K

Cue is an open-source AI copilot for the desktop: a small glass panel that floats above your windows, sees your screen, hears your microphone and the meeting audio, and suggests things in real time. Reach for it when you need to work out what to reply during a live call, when a coding problem is on screen and you want a solution from a screenshot, or when you want a recap of the conversation or a question answered about what is on screen. It offers smart assist, a what-should-I-say prompt, recap, free-form questions and a coding-problem solver on hotkeys. It runs on macOS (other side's audio needs 14.4+) and Windows 10 2004+/11, with a Linux build available, and speech recognition uses a bundled whisper.cpp. By default answers go through the paid publik API (per the README, requests cost 50% of the model's list price and prompts and screenshots pass through publik's servers), or you can use your own key for OpenAI, Anthropic, Google Gemini, Azure AI Foundry or a compatible endpoint. Install a prebuilt build from Releases or run from source (Node.js 22.12+, npm install && npm start). Licensed under GPL-3.0. Hiding from screen shares is best-effort, not guaranteed: on macOS 15.4+ the panel may be visible, and a phone camera always sees it. Use it only lawfully and with participants' consent; in proctored exams and job interviews it may break the rules. Not for analysing recordings after the meeting.