Heard
Reads your AI coding agent's activity out loud in plain language, so you can keep an eye on several Claude Code or Cursor sessions without staring at four terminal windows at once.
🔗 Visit Heard
Description
Running several AI coding agents in parallel means a wall of scrolling terminal text, most of which is routine noise — you only really need to know when something needs your permission, an error happened, or a task finished. Heard listens to that output and converts it into a short spoken summary, so you can work on something else, or step away, and still know when to pay attention.
Heard runs on macOS and works with Claude Code, Codex and Cursor, filtering agent output for "hard signals" like permission prompts and tool failures while summarizing routine progress at the project level — useful specifically when you're running several agents on different projects simultaneously. It includes a verbosity dial from full commentary down to errors-only, a phone-pairing feature (paid Heard Power tier) for remote monitoring and voice approval away from your Mac, and it's open source with a local session-state architecture rather than routing your code through a third-party server. The free tier covers personal use; Heard Power adds the remote/phone capabilities.
💬 Our review
The short version: if you run multiple AI coding agents at once and find yourself constantly tabbing between terminal windows just to check if anything needs you, Heard's spoken summaries let you offload that monitoring to your ears instead, for free.
It's closely related to Openbase, which goes further by adding voice input and remote control on top of voice output — Heard is more focused: it only gives your agents a voice, it doesn't let you talk back to them or approve changes remotely unless you pay for the Power tier's phone pairing. Against doing nothing and just watching terminal scrollback, Heard's real value is the hard-signal filtering — surfacing only what actually needs your attention rather than reading everything aloud, which would be worse than the terminal. Being open source and local-first (no third-party server relaying your code) is a meaningful trust signal for a tool that hooks into your coding sessions. Good fit for anyone running 2+ parallel agent sessions regularly; overkill if you only ever run one agent at a time and watch it directly.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit pour usage personnel · Heard Power (payant) ajoute le pairing téléphone pour monitoring à distance
Pros
Filtre les signaux importants (permissions, erreurs) du bruit de routine
Résumés au niveau projet, utile avec plusieurs agents en parallèle
Open-source, architecture locale (pas de serveur tiers relayant le code)
Gratuit pour l'usage personnel de base
Cons
Pas de contrôle vocal en retour (contrairement à Openbase)
Monitoring à distance par téléphone réservé au tier payant
Site retourne parfois 403 en accès direct
Utilité limitée si vous ne faites tourner qu'un seul agent à la fois