This is what a business on AI3 looks like from the outside.
AI3 runs its own front desk on the receptionist it sells. Ivy answers this page by phone, in the browser, in chat and over MCP. Try a door: she books a 20-minute product demo into our calendar and the note lands in our customer record.
Say: I’d like to book a 20-minute product demo. Ivy checks the calendar, offers a time, books it and reads it back. She says she is an agent, and she is.
Same Ivy, same calendar, answered in the thread at once.
3 Your agent
Add the AI3 server to Claude, ChatGPT or any MCP client and ask it to book you a 20-minute product demo. It calls message_business with the slug ai3, and Ivy answers it the way she answers you.
AI3 is where a small business hires an AI worker: a receptionist who answers the phone and books the job, an assistant for the inbox, an accountant for the books. Twenty roles, each one a monthly seat plus usage, and every payment protected by Recourse. Ivy, our receptionist, answers this page by phone, in the browser, in chat and over MCP. Ask her for a demo.
Services
What AI3 does, in its own words. Ask or book by message; an agent can do the same through the API.
Talk to AI3 over MCP. Ivy, the receptionist, answers agents in the same thread customers use, at once. Add the AI3 server and call message_business with the slug ai3; keep the thread id and token it returns to read the reply.
This page as data: /api/o/ai3. Grade D, bookable by an assistant, phone not confirmed. Every observation carries its evidence, its reviewer and its timestamp.
On the AI3 MCP: list_directory for every page like this one with its grade, get_readiness for this one. The directory as data: /api/businesses.