Clawdia Phone Bridge

Vapi voice assistant ↔ OpenClaw HTTP bridge. Near real-time phone calls with your agent.

The Clawdia Phone Bridge by @alejandroOPI is an HTTP bridge that connects Vapi — a voice AI platform for phone calls — to OpenClaw's gateway via WebSocket. The result: you can call a phone number and have a natural voice conversation with your OpenClaw agent. Vapi handles speech-to-text and text-to-speech on the call, while the bridge forwards questions to your OpenClaw instance and returns answers in near real-time. The architecture is elegantly simple. Vapi's voice assistant ('Clawdia') receives your spoken words, transcribes them, and makes a tool call to the bridge's `/ask` endpoint. The bridge forwards the question over WebSocket to the OpenClaw gateway, waits for the response, and returns it to Vapi, which speaks the answer back to you. The entire round-trip is fast enough for natural conversation flow. This project is significant because it gives OpenClaw a phone interface — arguably the most natural way to interact with an AI assistant. While Telegram and WhatsApp are great for text, there are moments when you just want to call your agent: while driving, cooking, or walking. The bridge runs as a lightweight Node.js service and can be exposed via Tailscale Funnel or ngrok for public access from Vapi's servers.

Tags: voice, vapi, bridge, phone

Category: voice

Tips

  • Use Tailscale Funnel (`tailscale funnel 3847`) for the cleanest way to expose the bridge to Vapi's servers — no port forwarding needed
  • Configure Clawdia's Vapi system prompt to be brief and direct — phone conversations need shorter responses than text chats
  • Set the bridge's GATEWAY_URL to your local OpenClaw WebSocket endpoint (default: ws://127.0.0.1:18789)
  • Test with the direct call format (`POST /ask` with `{"question": "..."}`) before configuring the full Vapi tool call flow
  • Consider adding a wake word or PIN requirement before the bridge processes requests to prevent unauthorized phone access to your agent

Community Feedback

HTTP bridge between Vapi voice assistant (Claudia) and OpenClaw. Phone → Vapi → Bridge → OpenClaw → Bridge → Vapi → Phone. Near real-time voice conversations with your agent.

— GitHub

Vapi voice assistant ↔ OpenClaw HTTP bridge. Near real-time phone calls with your agent. The missing phone interface for OpenClaw.

— OpenClaw Showcase

Frequently Asked Questions

Does Vapi cost money for phone calls?

Yes. Vapi charges per-minute rates for phone calls that include STT, TTS, and telephony costs. Check vapi.ai for current pricing. The bridge itself is free and self-hosted.

Can I use this with a different voice AI platform?

The bridge is designed for Vapi's tool call format, but the pattern is simple enough to adapt. Any voice platform that can make HTTP POST requests with a question and receive a text response can be wired up similarly.

What's the typical response latency?

End-to-end latency depends on your LLM provider speed and Vapi's STT/TTS processing. Typical round-trips are 2-5 seconds, which is acceptable for natural phone conversation with brief pauses.

Does it support multiple languages?

Yes. The original implementation uses Spanish (the author's preference), but Vapi supports dozens of languages for STT and TTS. Configure the language in your Vapi assistant settings.