Introduction
What Reception AI is, what ships in the package, how calls and chats flow through it, and where to start reading.
3 min readUpdated Oct 17, 2026Applies to Reception AI 1.0.0
Reception AI is a self-hosted AI receptionist for small businesses. It answers phone calls and website chat, answers everyday questions from the details you give it, books appointments into a real diary, captures leads and opens follow-up tickets. Every conversation, booking and lead lands in one workspace.
You can run it for a single business, or sell it to many businesses as a service with the included plans, Stripe billing and admin panel.
Tip
Want to see it before reading further? The live demo at receptionai.cc/login has one-click business-owner and platform-admin doors. It uses sample data and does not save changes.
What is in the package
| Area | What you get |
|---|---|
| Workspace | Receptionists built from 18 business templates or from scratch, a prompt editor and a visual call-flow builder, knowledge (documents, web pages, notes), opening hours and a booking diary |
| Channels | Phone lines on your own carrier account, a website chat widget, public forms, and a live voice demo on the marketing homepage |
| Operations | Conversations and transcripts, bookings, leads and tickets, contacts with CSV import and export, outbound campaigns, insights with CSV export |
| Connectors | 13 business connectors (Gmail, Google Calendar, Outlook, Slack, HubSpot, Salesforce, Shopify, Stripe, Zendesk, Notion, Calendly, MCP servers, plus Google Sheets export) and 7 AI provider connections |
| SaaS administration | Plans and allowances, usage metering, Stripe Checkout and customer portal, provider cost tracking, configuration, service health, roles and an exportable audit log |
| Source and tooling | Full TypeScript source, SQL migrations, admin CLI, a Dockerfile and a Railway infrastructure file |
It is built with TanStack Start (React 19), Nitro, Postgres, Better Auth and Tailwind CSS, and uses OpenAI Realtime for live voice and xAI for chat.
How a conversation flows
Reception AI talks to callers and visitors through two different engines. Knowing which one handles a conversation explains most of what you will read in these docs.
| Channel | Engine | Uses knowledge documents | Uses the call-flow builder |
|---|---|---|---|
Inbound phone call on Twilio or Plivo, with APP_URL set |
OpenAI Realtime, speech to speech | Only through the search_catalog tool |
No |
| Inbound call through a SIP trunk to OpenAI | OpenAI Realtime, speech to speech | Only through the search_catalog tool |
No |
Outbound Twilio call, or inbound Twilio without APP_URL |
Turn-based: carrier speech recognition, xAI model, text to speech | Yes | Yes |
| Website widget and workspace chat | xAI chat model | Yes | Yes |
| Homepage live voice demo | OpenAI Realtime over WebRTC, fixed demo persona | No | No |
Important
Live Realtime phone calls receive the receptionist's prompt, its guardrails, the open booking slots and its tools. They do not receive uploaded knowledge or the call-flow graph. Enable the search_catalog tool so a live call can look things up in your knowledge. Building a receptionist explains this in full.
What you need to run it
The short version, detailed on Requirements:
- A server that keeps a Node.js 22 process running (not serverless).
- Postgres for real use. An embedded in-memory database is included for trying it out.
- A domain with HTTPS.
- An OpenAI API key for live voice and an xAI API key for chat.
- A Twilio or Plivo account for real phone numbers.
Third-party usage (AI models, phone numbers and minutes, hosting) is billed to you by those providers and is not included in the purchase.
Where to go next
- Evaluating? Read Requirements, then Known limitations and the FAQ.
- Trying it locally? Quick start gets you running in about five minutes with no cloud services.
- Going to production? Follow Installation and deployment, then First admin and the admin panel.
- Selling it to businesses? Read Selling it as SaaS, including the licensing note.