VoicyClaw

Give OpenClaw agents a voice.

Talk in VoicyClaw, let OpenClaw run the agent, and hear the reply streamed back in real time.

VoicyClaw is the voice layer around OpenClaw bots. It gives you a web voice studio, realtime speech pipelines, and the connector path for bringing a real OpenClaw bot online.

What Works Today

• voice-first web studio for talking with a local demo bot or a real OpenClaw backend
• client and server speech paths that share the same turn-control UX
• server TTS adapters for Azure, Google Cloud, Tencent Cloud, and Volcengine
• hosted-style starter onboarding for workspace, project, and starter key flows
• OpenClaw plugin packaging under extensions/voicyclaw

Quick Start

cp config/voicyclaw.example.yaml config/voicyclaw.local.yaml
pnpm install
pnpm dev

Open http://localhost:3000.

Runtime Modes

• pnpm dev Source-mode local development. Starts the server, web app, and mock bot together.
• pnpm build && pnpm start:demo Built runtime. Uses the server dist, Next standalone output, and the built mock bot.
• pnpm test:e2e Builds first, then runs Playwright against the same built demo runtime.
• docker compose -f deploy/docker-compose.yml --env-file deploy/.env up --build Self-hostable Docker path.

Config Story

The main local config file is:

config/voicyclaw.local.yaml

That one YAML file is the public-facing source of truth for:

• app ports and public server URL
• auth mode
• storage driver
• demo bot defaults
• provider credentials

VOICYCLAW_CONFIG still exists as an advanced override, but the default repo story is one local YAML file.

Maps

• Developer Map
• Deployment Map
• Deploy Notes

Design Docs

• Project Definition
• OpenClaw Protocol
• Adapter Interface
• Gateway Bridge
• Conversation Backend Abstraction
• VoicyClaw Channel Plugin
• Artifact Standardization
• TTS Provider Roadmap
• Extreme Onboarding Flow
• Hosted Resource Model