PACT

One flat. One app. PACT explains confusing letters in your language, splits bills fairly and keeps chores on track: the only flat app you'll ever need.

Comment

Inspiration

Moving into a shared flat, especially in a new country, comes with a pile of paperwork nobody explains: council tax letters, energy bills, tenancy forms. Around 686,000 international students study in the UK, and many face these documents in a second language. Money is already tight (76% of students worry about making ends meet), and in shared homes, 30% of people argue over bills. The help exists, but it's scattered, confusing and rarely in your language. We wanted to build the app we wish we'd had on move-in day.

What it does

Pact is one hub for your whole flat, with six tabs: Docs, Bills, Chores, Roomates, Key Contacts, and Settings.

  • Docs: upload any letter or form and Pact summarises it in plain English, translates it into your language and sends key information straight to the bill tracker.
  • Read aloud: summaries can be read out loud for accessibility, in your chosen language.
  • Bills: split shared costs fairly between flatmates, allows for currency conversions for undertstanding, and allows for sending via WISE.
  • Chores: assign tasks and keep everyone on track with reminders.

It fits Wise's "Max Ease" brief by removing friction from everyday shared living: confusing forms become plain answers, no arguing over money owed and newcomers don't need local knowledge.

How we built it

  • Frontend: Next.js (React), Tailwind CSS and JavaScript, first scaffolded with Replit's AI Agent.
  • Backend: Next.js API routes that send documents to the Google Gemini 3.8 Flash Interactions API for summarisation, translation and email drafting.
  • Accessibility: the ElevenLabs text-to-speech API reads summaries aloud.
  • Email Knudge: We integrated the Resend API with a custom domain from GoDaddy to securely send the AI-drafted reminder emails directly to flatmates.
  • AI pair programming: as non-coders, we used DeepSeek and Claude in VS Code to write our backend API routes, debug environment variables and fix JSON parsing errors. We managed our own API keys securely in environment variables.

Challenges we ran into

  • We ran out of Replit credits mid-hackathon. Instead of stopping, we exported our code to GitHub, set up a local development environment in VS Code, and kept building.
  • Google deprecated the older Gemini models during the weekend, so our API calls started failing. We pivoted to the brand-new Gemini 3.8 Flash Interactions API.
  • Debugging API errors (like the Gemini model rejection and ElevenLabs JSON failures) by reading error logs and fixing the logic step by step.

Accomplishments that we're proud of

  • Going from zero coding experience to a working, accessible prototype in one weekend.
  • Integrating three live AI APIs (Gemini and ElevenLabs and ReSend) end to end.
  • Adapting to two major setbacks without giving up.

What we learned

How a web app is structured (frontend, backend API routes and external APIs), how environment variables keep API keys secret, how to use Git and GitHub, and how to debug by reading error messages. We also learned how to use AI tools responsibly and transparently as pair programmers.

What's next for Pact

  • Paying your share directly through Wise
  • Showing the original document side by side with the summary, so users can always check the AI's answer
  • Partnering with universities and letting agents to support new tenants

Built With

Share this project:

Updates

Submission history