Heard

What is Heard?

Heard is a free, open-source macOS utility that gives a voice to agentic coding tools like Claude Code, OpenAI Codex, and Cursor. Instead of watching a terminal scroll, developers hear concise spoken summaries of what their agent is doing, finding, or asking for, which makes it possible to step away, multitask, or monitor multiple sessions at once. The app runs as a lightweight menu-bar client, supports both fully local speech synthesis and premium cloud voices, and is built specifically to pronounce code, file paths, and developer jargon correctly rather than mangling them like generic text-to-speech.

Key Features

• Agent-to-Speech Narration

  Converts raw output from Claude Code, Codex, and Cursor into clear, human-sounding spoken updates in real time.

• Flexible Voice Engines

  Runs on free local Kokoro voices out of the box, or connects via bring-your-own-key to ElevenLabs, Anthropic, or OpenAI for higher-fidelity or custom personas.

• Built for Code

  On-Mac speech handling and a developer-specific vocabulary keep technical terms, function names, and file paths intact instead of garbled.

• Listening Modes

  Co-pilot, Companion, and Focus modes let users tune how much detail and interruption they want while an agent runs.

• Session Controls

  Pause and continue hotkeys plus automatic pausing during calls keep narration from clashing with meetings or other audio.

• Privacy-Scoped Pairing

  Explicit, scoped connections to agents ensure the app only listens to sessions the user has approved, avoiding ambient always-on capture.

Use Cases

• Hands-Free Monitoring : Developers can step away from the desk during long agent runs and still track progress, errors, or completion via audio.
• Multitasking During Builds : Engineers can keep working on other tasks or attend calls while spoken updates flag when an agent needs input or finishes a step.
• Accessibility-Friendly Coding : Audio narration offers an alternative interaction mode for developers who prefer or need less reliance on constant screen reading.
• Multi-Agent Oversight : Teams running several coding agents in parallel can distinguish and follow each session's status through distinct voice cues.
• White-Label Voice Layer : Product teams can embed Heard's narration API and SDK to add a voice co-pilot experience into their own agentic products.

FAQs