Descriptions:
Nate Herk demonstrates how to build a fully functional voice agent in roughly 15 minutes using Claude Code and ElevenLabs, without manually configuring anything inside the ElevenLabs dashboard. The central example is a custom agent trained on all 400 of Herk’s YouTube video transcripts — deployed as an embeddable widget on his website so visitors can ask questions in natural language and hear spoken responses sourced from that knowledge base.
The video explains the four components of every voice agent (persona/system prompt, voice, LLM backend, and tools), then walks through a live build of a second agent: a Cal.com appointment-booking assistant that checks availability and schedules calls. Claude Code handles API key wiring, system prompt drafting, tool configuration, and widget embedding autonomously after receiving a high-level description of the goal. Herk also covers why he prefers ElevenLabs for voice quality and interface clarity, and shows how to use bypass-permissions mode to let Claude Code execute the build end-to-end.
For developers exploring voice AI, the video is a practical end-to-end guide to the Claude Code + ElevenLabs stack, covering architecture concepts alongside reproducible implementation steps, including .env file setup and Cal.com OAuth authentication.
📺 Source: Nate Herk | AI Automation · Published May 04, 2026
🏷️ Format: Hands On Build
