OpenClaw plugin
Implementation shipped
The @papercup/openclaw-plugin package now ships a real SpeechProviderPlugin impl wrapping Kokoro. Loader-config docs are still TBD pending a live OpenClaw integration test — open an issue if you'd like to be the first to wire it up.
Why a plugin
OpenClaw is a 247k-star personal AI assistant with adapters for 25+ messaging platforms (Discord, Slack, WhatsApp, Telegram, Signal, Matrix, etc.). Its voice story today is platform-native speech (iOS Speech, Android SpeechRecognizer/TTS) — which works for the macOS/iOS/Android channels but doesn't apply to Discord.
Papercup's voice stack (Silero VAD + Whisper + Kokoro / MeloTTS / XTTS-v2) runs server-side and is exactly the missing piece for OpenClaw's Discord channel.
What the plugin will provide
- Server-side STT for incoming Discord voice
- Server-side TTS for outgoing Discord voice
- Reuses the same
@papercup/voice-stackpackage as the standalone bot - Plugs into OpenClaw's existing Discord adapter without forking it
The extension subagent system (Claude Code subagents in sandboxes) is also packageable as an MCP plugin — OpenClaw consumes MCP-standard tools, so that part is mostly free.
Track this
Watch the @papercup/openclaw-plugin package in the GitHub repo for progress.