Works with OpenClaw
Partial · Universal wrapper
The open source Heard engine can sit around an OpenClaw terminal process with its universal heard run wrapper, which gives the process a spoken output layer. heard run is part of the open source engine, not the managed Mac app, and there is no dedicated OpenClaw adapter today.
Heard labels this integration as “Universal wrapper.” That label describes how Heard receives the work rather than treating every tool as if it had the same native connection.
- Current status
- Partial
- Integration path
- Universal wrapper
- Desktop platform
- macOS 14 or later
- Talk back
- Heard PowerSupport depends on the underlying session path.
- Open source option
- Apache 2.0 for the open source engine
Setup
Set up Heard for OpenClaw
-
Install Heard
Install the open source Heard engine from github.com/heardlabs/heard.
-
Connect OpenClaw
Launch the OpenClaw command through
heard run.Open source engineheard run <command> [args...] -
Finish setup
Pick the open source engine’s listening mode (Co-pilot, Companion, or Focus) and confirm which messages your workflow exposes.
What you will hear
Narration is a filter, not a transcript
Heard looks for progress, blockers, decisions, failures and useful results. How much routine output is spoken depends on the open source engine's listening mode.
heard run wraps Hermes and OpenClaw so Parrot can narrate them.Terminal based output capture
Concise spoken updates rather than raw line by line reading
Use beside first class Claude Code and Codex sessions
Narration here comes from the open source engine, which has three listening modes: Co-pilot, Companion and Focus.
Limits
Know the boundary before you leave the screen
A reliable eyes off workflow starts with an accurate support boundary. Run a short test task, confirm that questions and final results reach Heard, and keep the original tool available for interactions the integration cannot route.
- There is no dedicated OpenClaw adapter today.
- The wrapper can only interpret output available to the terminal process.
- Actions that require an OpenClaw specific interface may still need the original UI.
Questions
Frequently asked questions
Does Heard work with OpenClaw?
The open source Heard engine can sit around an OpenClaw terminal process with its universal heard run wrapper, which gives the process a spoken output layer. heard run is part of the open source engine, not the managed Mac app, and there is no dedicated OpenClaw adapter today.
What kind of OpenClaw support does Heard provide?
The current support level is universal wrapper. The page above names the exact setup path and limitations so “works with” does not hide the difference between a native adapter, a wrapper, and a terminal host.
Does Heard block OpenClaw while it speaks?
No. Heard processes narration asynchronously. The coding agent can continue working while speech is generated and played.
Can I use Heard with several agents at once?
Yes. Heard keeps session context separate, gives attention to blockers and decisions, and can roll lower priority activity into a shorter spoken brief.
Can I talk back to the agent?
Talking back is part of Heard Power: press right ⌘ and tell Parrot what to send, and it goes to the session you name. Availability can depend on the integration path and whether the underlying tool accepts the response through the active session.
What leaves my Mac?
The open source engine can use local Kokoro speech and your own keys. Managed narration can send a compact narration snippet to configured service providers. Heard does not retain source code, terminal output, narration text, or generated audio after a hosted request completes.
Ready when your agents are
Hear the work without watching every screen.
Start with 14 days of free Heard: the full Power experience. No card required. The open source engine remains free.
Download Heard for Mac macOS 14 or later · Claude Code and Codex supported
