github.devotd.
NewAI agents

louis-cfm/coucou v0.4.0 — standalone dock app and voice features

Introduces the changed macOS app structure of louis-cfm/coucou v0.4.0 and the newly added voice input and call transcription features.

Repositorylouis-cfm/coucou ↗

A tiny friend in your Mac's notch and on your iPhone that keeps an eye on your AI coding agents: Claude Code, Codex, Cursor, Gemini CLI, Antigravity and more. Approve from the notch or your Lock Screen.

stars
★ 5.4K
Forks
948
Language
Swift
License
MIT
Latest release
v0.4.0
Read original source ↗View on GitHub ↗

louis-cfm/coucou v0.4.0 is a new release of the Swift app that shows coding agent status. Starting with this version, it uses a standalone dock app and separate windows instead of the Mac notch. The left window holds the Flow voice feature, and the right window holds notch-area settings. According to the repository description, it is a tool that handles sessions of tools such as Claude Code, Codex, Cursor, and Gemini CLI.

The app shows agent progress and lets you respond to permission requests and AskUserQuestion from the notch or the lock screen. Other agents can be connected by putting a coucou_agent tag in the hook payload, and the integration method is documented. v0.4.0 also adds voice input that pastes what you said at the cursor, plus call recording transcription and summarization. Voice input uses the right ⌥ key as its default trigger and runs NVIDIA Parakeet through FluidAudio to recognize 25 languages offline. The model download is about 650MB, and this feature is Apple Silicon only. The call feature produces a transcript and summary per speaker, key points, and to-dos, and deletes the audio once the transcript is written.

coucou is an open-source app under the MIT license, but the name and graphics belong to the author. The provided materials contain no performance benchmarks or model evaluation results, so the processing performance of the features is hard to compare. The strengths the materials support are agent integration and the usage flow, and there is not enough evaluation information to use it as a performance-driven reason to choose it.