openclaw-voice-gpt-realtime

What This Skill Does

The openclaw-voice-gpt-realtime skill empowers your OpenClaw agent to perform natural-sounding, real-time phone calls using OpenAI's advanced speech-to-speech engine. By leveraging the Realtime API, this skill eliminates the traditional latency associated with separate speech-to-text and text-to-speech pipelines, resulting in conversational response times between 200ms and 300ms. The agent acts as an autonomous caller, capable of navigating IVR (Interactive Voice Response) menus, identifying voicemail, and handling complex interactions like appointment booking or inventory checks. It essentially puts a human-like voice interface directly into your terminal or automation workflow, providing structured JSON results once the call concludes.

Installation

To begin, ensure you have your Twilio account credentials and an OpenAI API key with Realtime API permissions. First, install the skill via the command clawhub install openclaw/skills/skills/connorcallison/openclaw-voice-gpt-realtime. Once installed, edit your openclaw.json or plugin settings to include your twilio.accountSid, twilio.authToken, fromNumber, and openai.apiKey. Because this skill requires a webhook to receive voice events from Twilio, you must expose a public URL that tunnels traffic to your local server (port 3335). You can use tools like Cloudflare Tunnel or ngrok for this. After configuration, execute openclaw voicecall-rt status to verify connectivity.

Use Cases

This skill is perfect for automating mundane administrative tasks that require phone interaction. Common use cases include booking restaurant reservations, confirming availability of products at local retailers, scheduling service appointments for homes or businesses, and querying customer support lines for account updates. It is particularly useful for agents managing personal or business logistics where digital APIs are unavailable.

Example Prompts

1. "Call the local florist at +15550123456 and ask if they can deliver a bouquet to my office by 4pm today."
2. "Please call my dentist at +15559876543 and reschedule my appointment from Tuesday to next Friday morning."
3. "Call the tech support line at +15551112222 and find out why my current subscription status is showing as inactive."

Tips & Limitations

Be mindful of costs, as each minute is billed by both OpenAI and Twilio, averaging around $0.31 per minute. The system is configured for 5 concurrent calls by default; increase this only if your hardware can handle the processing load. Always enable debug: true during your initial testing to monitor latency metrics and review call transcripts, though be aware that these recordings may contain sensitive information. The skill includes built-in safety guardrails to prevent malicious identity spoofing, ensuring all calls adhere to ethical usage standards.