Inspiration

One of my friends runs a restaurant and is always on-call. Since he runs a small business, he does not have the funds to hire a receptionist to take phone calls and orders. I built Wattle so that an AI agent is able to take phone calls and record and book reservations for the restaurant. It also covers offers a web widget or inline form which users can embed on their website so inbound queries can go to a unified inbox.

What it does

Wattle is a voice AI agent that is integrated with Google Sheets and Twilio. A user is able to purchase a number via Twilio for the agent to receive calls from. With Google Sheets it is able to collect customer details and store them. It also has a unified inbox and CRM that keeps track of customers sending queries or any other inbound queries via the web widget and is able to summarise calls made with the voice AI agent. The UI is designed with a setup agent to allow a user to quickly get their agent up and running,

How we built it

This was built using Codex and GPT 5.6. GPT 5.6 accelerated building all the features and was pivotal in configuring the app for production using the various MCP using Stripe, Twilio and Porkbun MCPs. I would write and generate a plan first for a particular feature, review it and then implement it. I enjoyed using the end-to-end Chrome to verify the frontend and different pages in the app. Our tech stack was NextJS + Supabase and Google Cloud. This tech stack had MCPs which cut down hours configuring GCP instances and figuring out what migration or table to build in Supabase.

Challenges we ran into

I was challenged with the lack of time to fully implement all the integrations. However, the majority of the app works.

Accomplishments that we're proud of

I am proud that we built the app in a working prototype in a week. Even though we were pressed for the time, a large amount of features were built and worked on including the purchase flow for the Twilio numbers and the node editor for the voice AI agent which allows a user to make the voice agent deterministic.

What's next for Wattle

We aim to get some paying customers afterwards.

Built With

Share this project:

Updates