GoDo AI

A prototype app that suggests things to do in the evening, based on how busy someone's day has been, their budget and where they are. We built it as a team of three.

Agents
Tool calling
Claude API
React Native

Context: UCL Machine Learning for Business, group coursework
Date: March 2026
Team: Three, with Mahira Malikzada and Olena Brylinska
Stack: React Native, Claude API, Google Calendar, Eventbrite, Ticketmaster, Skiddle, Google Places

Two layers

A rules layer handles the parts that should not be left to a language model. It scores how full the day has been, maps that to how demanding a suggestion should be, and applies budget and time filters. Claude then works within those limits to choose suggestions and explain them.

Two modes

Daily picks ranks events by budget, free time and distance and returns five. Scout is a chat agent that can call event and places APIs over up to three rounds and streams its reply to the app.

Every venue is checked against a live API before it is shown, so the app does not suggest places that do not exist. Calendar access is read-only and limited to free and busy times.

Alongside the prototype we produced a pitch deck and a recorded walkthrough.

Screenshot of the GoDo AI app showing today's suggested events with interested and not interested buttons. — The daily picks screen in the prototype.

Slide describing the rules-based layer and the Claude layer of the GoDo AI architecture. — The two-layer design, from the pitch deck.

Slide comparing Daily mode, which ranks top picks, with Scout mode, an agentic chat. — The two recommendation modes, from the pitch deck.