Catava
An AI receptionist that answers every call, chat, email and WhatsApp message, and books the appointment while it's at it.
The short version.
Catava is 7th Pillar's own SaaS product: a receptionist that never takes a day off. It picks up the phone, talks to visitors on the website by voice, replies to chat, email and WhatsApp, and does real work in each conversation: it checks the calendar, books and reschedules appointments, takes orders from the catalogue, and leaves the team a neat summary and a follow-up task.
It started as a voice receptionist for a medical clinic and grew into a general platform for small businesses. We tested it end to end against three very different sample businesses, a clinic, an Italian restaurant and a vacation rental, so the same product has to handle appointments, table orders and stay enquiries equally well.
What was broken.
A missed call is a lost customer. Small businesses get enquiries from four or five directions at once, usually while the owner is busy with someone in front of them. Voicemail doesn't book appointments, and a full-time receptionist covering phone, chat, email and WhatsApp is out of reach for a five-person clinic or a family restaurant.
- Calls after hours and during rush periods go to voicemail and never call back.
- Chatbots answer FAQs but can't check a slot, take a booking or place an order.
- Customer history is scattered across the phone log, the inbox and WhatsApp.
From idea to production.
- Jan 2026
Voice receptionist for a clinic
A working phone and browser voice agent that could look up services, check availability and book appointments.
- Mar 2026
Multi-channel platform
Chat, email and WhatsApp joined voice, sharing the same tools and customer memory. The 4-step setup wizard, catalogue, orders and billing modules were added.
- Apr 2026
Moved to the edge
The web platform moved to Cloudflare's global network for faster response times worldwide.
- May 2026
Sounds like a person
The agent's speaking style was rewritten so it uses natural phrasing, confirms details the way a person would and never reads out lists like a robot.
What we built.
One AI receptionist, five channels, one memory. Catava's agents share the same catalogue, calendar and customer history whether the customer calls, types or emails. A business describes itself in a sentence, pastes its website or uploads a photo of its menu, and Catava drafts the full service catalogue with prices, sets up an agent personality and voice, and hands over embed codes for the website. Most businesses can go live the same afternoon.
- Step 1Customer calls or messages
- Step 2Agent understands the request
- Step 3Checks catalogue, calendar and history
- Step 4Books, orders or creates a task
- Step 5Summary lands in the dashboard
What we delivered.
Five channels, one brain
Phone, website voice, chat, email and WhatsApp all use the same tools, knowledge and customer memory. A customer can start on WhatsApp and finish on a call without repeating themselves.
Setup in four steps
Describe the business, paste a website address or upload a menu photo. AI drafts services, products, prices and the agent's personality. The owner reviews, picks a voice, and copies the embed code.
Real bookings, not messages
A month, week and day calendar with reschedule, confirm and no-show handling, kept in sync with Google Calendar and Calendly.
Ordering inside the chat
Product cards with image carousels let customers browse and order without leaving the conversation. Prices can depend on the customer's answers, such as number of rooms or guests.
Customers that build themselves
Every conversation creates or updates a customer record, with the full history across channels, so the next agent (human or AI) has context.
A knowledge base that listens
Upload PDFs, audio and video. Catava transcribes and extracts them so every agent can answer from the business's own material.
Tasks the team can act on
Callbacks, follow-ups and payment reminders land on a Kanban board, created by the agent during the conversation.
Pay for what you use
Plans come with dollar-based credits that never expire. Every AI call is metered, so usage and cost are always visible.
A closer look.
Intelligent, and trustworthy.
Real-time voice
Speech is transcribed as the customer talks, the agent decides what to do, and a natural voice replies, fast enough to feel like a normal phone call.
Same skills everywhere
The voice agent and the chat agent share one set of actions: look up a service, find a free slot, book, place an order, search the knowledge base, create a task.
Memory across channels
Before replying, the agent sees what the customer said on other channels, so the chat agent knows about last week's email.
Built-in cost control
Every AI call checks the balance first, records its true cost afterwards, and warns the business before credits run low.
The stack.
- Product
- Next.js web app, embeddable voice and chat widgets
- Voice
- Pipecat pipeline, Deepgram speech recognition, Cartesia voices, Daily WebRTC
- AI
- OpenAI GPT-4o and GPT-4o mini, Gemini for video
- Data
- Supabase Postgres with vector search, per-business data isolation
- Integrations
- Twilio phone numbers, Google Calendar, Gmail, Calendly, WhatsApp Business, Dodo Payments, Resend
The numbers.
Build proof
Expected impact for users
Verified from project repositories, test runs and release records. Projected modelled estimates of user impact, not measured client results.
Got an idea? Let's ship it.
Tell us the workflow that's costing you customers or hours. We'll come back with a prototype plan, a timeline in weeks, and a fixed quote.