Frequently asked questions
What does Heard do?
Heard is a voice layer for coding agents. It turns agent activity into spoken progress, decisions, blockers, and completed work. Heard Power also lets you answer by voice and dictate anywhere your cursor can type.
What is a voice layer for coding agents?
A voice layer sits alongside tools like Claude Code and Codex and turns their activity into speech: progress, decisions, blockers, questions, and results, spoken as they happen. Heard is the voice layer for coding agents on macOS: it listens to agent events locally, judges what deserves airtime, and narrates in a natural voice, with hands-free replies and codebase Q&A on Heard Power.
How do I make Claude Code talk?
Install Heard, enable the Claude Code integration, and start a session: progress, questions, and results are read aloud automatically. Setup takes a few minutes; see the Claude Code guide.
Can I get notified when my coding agent finishes or needs me?
Yes, that is Heard’s core job. A completed task, a failure, or a question waiting on you is spoken out loud, and the notch pops a card you can act on. Focus mode narrows this to interruptions only, so you can work on something else until an agent actually needs you.
Does Heard read every tool call aloud?
No. Heard judges what deserves airtime. Routine tool activity can stay quiet while failures, choices, approvals, and useful results are narrated.
Will Heard slow down my coding agent?
No. Supported hooks send events locally without waiting for narration. Speech generation and playback happen asynchronously.
Which coding agents does Heard support?
Claude Code, Codex CLI, and the Codex app have first class support. Other compatible command line agents can run through the heard run universal wrapper.
Does Heard support Cursor?
Heard can wrap a Cursor command line workflow with heard run. The Cursor editor itself does not have a first class native Heard adapter today.
Is Heard open source, and what is the difference from the app?
The Heard engine is open source under Apache 2.0 at github.com/heardlabs/heard. Clone or fork it to run Heard free forever, with local Kokoro speech or your own provider keys, built and updated by you. The downloadable macOS app is the official managed build of Heard: it adds hosted voices and managed narration on Pro, and the voice-input suite on Power: push to talk dictation, spoken replies, ambient input, Heard Parrot, and iPhone pairing. Those Power features ship only in the app; they are not in the open source engine.
Can Heard run locally?
Yes. The Apache 2.0 engine can use local Kokoro speech and neutral local templates. You can also bring your own voice and narration provider keys.
Does my code leave my Mac?
The answer depends on the services you choose. Local Kokoro speech stays on your Mac. Hosted narration receives compact event context and relevant response text. Cloud voice providers receive only the text Heard decided to speak. See Privacy and data.
Does microphone audio leave my Mac?
No hosted service is required for dictation. Heard Whisper performs speech recognition on device.
What are the listening modes?
Co pilot gives short signposts for screen based work. Companion gives fuller briefings when your eyes are elsewhere. Focus speaks only for important interruptions such as approvals, blockers, failures, and decisions.
Can Heard follow several agents at once?
Yes. Heard ranks active sessions by importance, gives sessions distinct voices, and summarizes background activity. You can pin one session when it needs your full attention.
What is the difference between Heard Pro and Heard Power?
Pro is the listening tier: premium hosted voices and managed narration with zero key setup. Power adds the talking half: push to talk dictation, spoken replies, ambient input, Heard Parrot, and iPhone pairing. See Heard Pro and Heard Power.
What is Heard Power?
Heard Power adds spoken replies, push to talk dictation, ambient voice input, and browser based iPhone pairing to the managed app.
Does Heard talk over my meetings?
No. When another app starts recording from the microphone, Heard holds its narration and catches you up after the call. Known meeting apps also mute narration in ambient mode.
Can I teach the dictation my project’s words?
Yes. Add custom vocabulary so Heard Whisper gets your project names, teammates, and jargon right. From Heard 1.1.53 the vocabulary syncs with your account and survives updates.
Is there a free trial?
Every new account starts with a 14-day trial of the full Power feature set, no card required. Invites stretch it further: sign up through a friend’s link for a 30-day trial, and earn a free week of Pro for each friend who activates through yours. See Plans.
Can I ask Heard about my codebase?
Yes. Hold the talk key and start with your narrator’s name (“Jarvis”, “Aria”, “Friday”, “Atlas”) or “Hey Parrot”: Heard Parrot runs a read-only dig through the active repo and answers out loud, with follow-ups and a spoken “catch me up” recap across sessions. See Heard Parrot.
What is the notch?
Heard’s visual surface under your camera: active sessions as lanes, questions as focused cards you can answer in place, and a done card when work finishes. See The notch.
Is the notch like the iPhone’s Dynamic Island?
Yes, that is the closest comparison: a Dynamic Island for your Mac, built around coding agents. It idles as a compact pill in the Mac’s notch area, expands with live session activity, pops when a task completes, and collapses when nothing needs you. It works on Macs with or without a physical notch.