Скачать или смотреть Gemini 3 + ElevenLabs Agent + Custom MCP for Roller API

Haven’t posted in a bit, but this one’s a fun build. In this video I walk through an AI voice agent I’m creating for a mini-golf venue in the UK. The agent runs on ElevenLabs and handles real bookings end-to-end — checking availability, creating reservations, and confirming payment — all through a custom MCP server I built to integrate Roller’s APIs into the ElevenLabs agent toolchain.

I’ll show you:

Why I like the ElevenLabs agent interface

The DevTools integration hiccup I ran into

How I set up a custom MCP server as a “servable” HTTP tool host

Live booking + payment flow (Roller + Stripe)

Email confirmation + SMS limitations with UK/US phone rules

How I parse Roller’s availability response so the agent books the right slots

The idea of training agents around the “80% most common caller goal”

If you’re building voice agents that need real operational control (bookings, CRMs, payments, etc.), this should give you a solid template.

Chapters / Timestamps
0:00 Intro + project overview
0:21 ElevenLabs interface + DevTools issue
0:28 Building a custom MCP server for Roller
1:08 Hosting MCP on Google Cloud
1:24 Live booking demo (email + payment check)
4:17 MCP tool calls + availability logic
5:50 Booking creation + Stripe + Twilio setup
6:46 Product/package mapping in Roller
8:18 Payment verification + custom HTML callback
9:07 Designing agents for the “80% use case”
9:46 Wrap-up

Tech stack: ElevenLabs Agents • Roller API • MCP custom server • Google Cloud • Stripe • Twilio

If you want a deeper follow-up on the MCP server setup / Google Cloud deployment, drop a comment and I’ll make a dedicated video.

👍 Like / subscribe if this helped
💬 Questions welcome in the comments

#AI #VoiceAgent #ElevenLabs #MCP #RollerAPI #Stripe #Twilio #ConversationalAI #Automation #GoogleCloud #AIAgents

https://everydaysoftware.ai/