agent-agnostic · on-device · open source
Callout runs silently while your AI agents work. It speaks up when they need a decision. You reply out loud, from anywhere.
Quick install
curl -fsSL https://raw.githubusercontent.com/TomasPhilippart/callout/main/install.sh | sh
Requires macOS 13+ or Linux. Build from source on GitHub.
How it works
Your agents call a local HTTP API. Callout speaks the question. You answer out loud. The agent continues.
01
Your agent hits a decision point
Instead of printing to a terminal or opening a dialog, it sends one request to Callout and waits. That's the whole integration.
02
Callout speaks up
It says the agent's name, then the question, then the options. You always know who's asking, even with three agents running at once.
03
You answer from wherever you are
Hold the hotkey, say "B" or "keep them", release. You don't need to look at the screen. Say it from across the room if you want.
04
The agent gets the answer and moves on
The request resolves with your reply. No polling, no callback URL, no window to switch to. The agent just continues from where it was waiting.
Features
Private by design. Agent-agnostic. Built for developers who run multiple agents concurrently.
Hold to record · release to answer
Experience
Hold a key, say your answer, release. No window switching, no clicking, no typing. Works across every app and every space on your machine.
Privacy
Voice recognition and speech synthesis run entirely on your machine. Nothing is sent to the cloud. Built in Rust: fast startup, low memory, single binary.
Awareness
Each agent registers with a name. Callout announces who's speaking and tells you who's blocked, even with three agents running at once.
Install Callout, connect your first agent, and answer your first question. All in under a minute.
curl -fsSL https://raw.githubusercontent.com/TomasPhilippart/callout/main/install.sh | sh
Requires macOS 13+ or Linux. Build from source on GitHub.