Heard - Give Claude Code and Codex a voice
Heard is a macOS voice layer for your agentic workflows. It connects to Claude Code, Codex, and Cursor and turns their output into intelligent summaries you can hear. Full commentary when your eyes are elsewhere, or silence until something errors or needs a decision. With agents run in parallel, Heard summarizes at the project level, so you hear where the work stands, not five terminals talking over each other. Pair your phone and it comes with you. Open source, free for personal use.
Replies
AISA (aisa.to) AI Fluency Assessment
really smart approach to the feedback loop problem with agentic coding. half the time I'm running Claude Code I'm doing something else entirely and only check back when it's done — having audio summaries of what happened means you catch errors or odd decisions way earlier instead of reading through the whole diff after the fact. does it differentiate between routine completions vs things that actually need your attention?
Heard
@ozandag yes you can switch between modes and have settings to talk with continuous summaries or only when things need your attention!
Terminal Candy
Congrats on the launch Kelly! The thing I keep hitting is not knowing whether an agent is actually stuck waiting on me or just still working. Does Heard speak up only when it needs an answer from you, or does it narrate as it goes?
Heard
@terminal_candy hey peter you can pick different modes as settings and it does both!
@itskellysun, this is awesome! Any chance you can share an invite code for the Power beta?
Heard
@whyyasgwhy only for our PH fans :) go to: https://heard.dev/power and code: friendsofheard!
I like the “useful update, not a transcript” framing. I’ve definitely lost time babysitting parallel agent terminals; curious how you decide when to stay silent vs interrupt for a decision?
Great product. When are you deploying for Windows ?
Heard
@iamme if we get enough demand! :)
@itskellysun Well, am demanding on behalf of other Windows users.
Macaly
#1 well deserved 🥇 talking to claude code instead of typing just makes sense
The part that got me is treating the session as having memory instead of judging each line in isolation, that's the actual hard problem here. most "smart notification" type tools just fire on keyword or event triggers and end up either too noisy or too quiet depending on the day
Curious about the failure case, when Heard is wrong about something being routine and stays quiet, how do you find out about that after the fact. is there a way to go back and ask it what happened while you were away, or once it decides not to speak is that moment just gone
Heard
@elhart05 it keeps a record of all the summaries and you can go back and ask it!
The multiple parallel agents talking over one another has been a huge problem for me too. Pretty excited to try this app out. I think it would make managing multiple simultaneous agents a lot easier. Nice work. :)
Heard
@hassanaziz0012 thanks man! would love your feedback once you give it a spin :)
Voice as the interface to a coding agent is more interesting than it first sounds, because the hard part isn't dictation, it's knowing when to interrupt. If I'm talking while the agent is mid-task, does it queue what I said, or does it treat it as a course correction and abandon the current step? Curious where you landed on that.
Meridian
Congrats on the launch. How does Heard handle conflicting updates across parallel agents? If one agent finishes successfully while another discovers a change that invalidates its work, does the project-level summary surface that dependency, or does each session remain isolated?
Heard
@akarsh_hegde i believe it would think it's important enough to surface that as dependency!
Meridian
@itskellysun That makes sense. Surfacing the dependency is the critical part. Would Heard also explain why the work may be invalidated, or only flag that the sessions conflict?