ClawKit Logo
ClawKitReliability Toolkit
Back to Registry
Official Verified

voice-agent-builder

Build and manage Voice AI agents using Vapi, Bland.ai, or Retell. Create agents, configure voices, set prompts, make outbound calls, and retrieve transcripts. Includes platform comparison guide. Use when building phone agents, IVR systems, or voice-first customer service.

skill-install — Terminal

Install via CLI (Recommended)

clawhub install openclaw/skills/skills/aiwithabidi/voice-agent-builder
Or

Voice Agent Builder

Build, configure, and manage Voice AI agents. Supports Vapi (primary), Bland.ai, and Retell platforms.

Quick Start

export VAPI_API_KEY="your-vapi-api-key"

# Create a voice agent
python3 {baseDir}/scripts/vapi_agent.py create-agent '{"name":"Sales Agent","firstMessage":"Hi! How can I help you today?","systemPrompt":"You are a helpful sales assistant for Acme Corp."}'

# Make an outbound call
python3 {baseDir}/scripts/vapi_agent.py call '{"assistantId":"asst_xxx","phoneNumberId":"pn_xxx","customer":{"number":"+15551234567"}}'

# List agents
python3 {baseDir}/scripts/vapi_agent.py list-agents

# List calls
python3 {baseDir}/scripts/vapi_agent.py list-calls

Platform Comparison

FeatureVapiBland.aiRetell
Best ForCustom agents, dev-friendlySimple outbound campaignsEnterprise, low latency
Latency~800ms~500ms~500ms
Languages100+30+30+
Custom LLM✅ Any OpenAI-compatible✅ Limited✅ Via API
Phone NumbersBuy/importBuy/importBuy/import
Pricing$0.05/min + provider costs$0.09/min all-in$0.07-0.15/min
WebSocket
Knowledge Base✅ Built-in
Transfers

Recommendation: Start with Vapi — most flexible, best docs, largest community. Use Bland for simple high-volume outbound. Use Retell for enterprise low-latency needs.

See {baseDir}/scripts/voice_comparison.md for detailed breakdown.

Agent Creation Workflow

1. Choose a Voice

Vapi supports multiple TTS providers:

  • ElevenLabs — Best quality, most natural (recommended)
  • PlayHT — Good quality, lower cost
  • Deepgram — Fast, good for real-time
  • Azure — Enterprise, many languages

2. Configure the Agent

{
  "name": "Appointment Setter",
  "firstMessage": "Hi! This is Sarah from Dr. Smith's office. I'm calling to help you schedule your appointment.",
  "systemPrompt": "You are Sarah, a friendly appointment scheduler...",
  "voice": {
    "provider": "11labs",
    "voiceId": "21m00Tcm4TlvDq8ikWAM"
  },
  "model": {
    "provider": "openai",
    "model": "gpt-4o",
    "temperature": 0.7
  },
  "endCallFunctionEnabled": true,
  "maxDurationSeconds": 300,
  "silenceTimeoutSeconds": 30
}

3. Prompt Engineering for Voice

Voice prompts differ from text. Key principles:

  • Keep responses SHORT — 1-2 sentences max per turn
  • Be conversational — use filler words naturally ("Sure thing!", "Got it!")
  • Handle interruptions — voice agents get cut off, design for it
  • Confirm understanding — repeat back key info (names, numbers, dates)
  • Include fallback — "I didn't catch that, could you repeat?"

4. Phone Number Setup

# List available phone numbers
python3 {baseDir}/scripts/vapi_agent.py list-phones

Metadata

Stars4473
Views0
Updated2026-05-01
View Author Profile
AI Skill Finder

Not sure this is the right skill?

Describe what you want to build — we'll match you to the best skill from 16,000+ options.

Find the right skill
Add to Configuration

Paste this into your clawhub.json to enable this plugin.

{
  "plugins": {
    "official-aiwithabidi-voice-agent-builder": {
      "enabled": true,
      "auto_update": true
    }
  }
}
Safety NoteClawKit audits metadata but not runtime behavior. Use with caution.

Related Skills

agent-memory

Full AI agent memory stack — Mem0 unified memory engine with vector search (Qdrant) and knowledge graph (Neo4j), plus SQLite for structured data. Complete setup script and tools. Give your OpenClaw agent a real brain with semantic recall, entity relationships, and structured storage.

aiwithabidi 4473

skill-auditor

Security audit and quarantine system for third-party OpenClaw skills. Use when evaluating, reviewing, or installing any skill from ClawHub or external sources. Automatically triggered before any skill installation.

aiwithabidi 4473

Model Intel

Live LLM model intelligence and pricing from OpenRouter

aiwithabidi 4473

deep-search

Multi-tier Perplexity search with Langfuse tracing. Three depth levels: quick (sonar), pro (sonar-pro), deep (sonar-reasoning-pro). Supports focus modes: internet, academic, news, youtube, reddit. Returns AI-synthesized answers with citations. Use for research, comparisons, market analysis, fact-checking. Triggers: search, research, look up, find out, compare, what is, deep search, web research.

aiwithabidi 4473

amplitude

Amplitude product analytics — track events, analyze user behavior, run cohort analysis, manage user properties, and query funnel/retention data via the Amplitude API. Understand product usage, measure feature adoption, and analyze user journeys. Built for AI agents — Python stdlib only, zero dependencies. Use for product analytics, user behavior tracking, funnel analysis, retention analysis, and cohort segmentation.

aiwithabidi 4473