BridgeSpeak: skill pack for Hermes Agent

Cross-agent skill that lets Claude Code, Hermes and OpenClaw speak aloud through OpenAI gpt-realtime-2

What BridgeSpeak does

BridgeSpeak is a cross-agent skill from BridgeMind that lets Claude Code, Hermes and OpenClaw agents speak aloud. The agent runs bash speak.sh with text, a Python WebSocket client connects to OpenAI's gpt-realtime-2, streams 24 kHz mono pcm16 audio, wraps it as WAV and plays it through the system's native audio player. Voices such as marin and cedar are named in the README.

The skill follows the agentskills.io standard and tells the agent when to speak and when not to. It includes speak.py, a POSIX speak.sh wrapper for macOS and Linux and a speak.ps1 wrapper for Windows. For Hermes it is copied into ~/.hermes/skills/voice/bridgespeak, and Hermes passes OPENAI_API_KEY through by reading the skill's required environment variables metadata. The only Python dependency is websockets, and Linux needs one of paplay, aplay or ffplay.

Key features

speak.sh and speak.ps1 entry points for macOS, Linux and Windows
Python WebSocket client for OpenAI gpt-realtime-2
Plays audio through the system's native player
Follows the agentskills.io skill standard
Guidance for the agent on when to speak and when not to
Separate install steps for Claude Code, Hermes and OpenClaw

When to use it

Hearing a spoken build summary such as test results while away from the screen
Leaving a Hermes Agent running and getting audio announcements
Narrating progress when working on a second monitor

Who it is for: Developers who run coding agents, including Hermes Agent, and want spoken status updates.

How it fits with Hermes Agent

It supports Hermes Agent alongside Claude Code and OpenClaw, with a documented copy into ~/.hermes/skills/voice and OPENAI_API_KEY passed through skill metadata.

How to install BridgeSpeak

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

mkdir -p ~/.hermes/skills/voice
cp -r skills/bridgespeak ~/.hermes/skills/voice/bridgespeak
cp -r scripts ~/.hermes/skills/voice/bridgespeak/scripts
python3 -m pip install --user websockets

Requirements: An OPENAI_API_KEY, the Python websockets package and a system audio player (paplay, aplay or ffplay on Linux)

Note: Speech depends on OpenAI's gpt-realtime-2 service, so an OpenAI API key is required.

FAQ

What is BridgeSpeak?

BridgeSpeak is a skill that lets coding agents speak aloud. It sends text to OpenAI's gpt-realtime-2 over a WebSocket and plays the returned audio on the user's speakers.

Does BridgeSpeak work with Hermes Agent?

Yes. The README includes a Hermes install that copies the skill into ~/.hermes/skills/voice/bridgespeak. Hermes reads the skill's metadata and passes OPENAI_API_KEY through.

What do I need to run BridgeSpeak?

You need an OpenAI API key, the Python websockets package and a system audio player. macOS and Windows have built-in playback; Linux needs paplay, aplay or ffplay.