Unbound

Inspiration

London has over 685,000 international students in the UK. Many struggle silently — not because help doesn't exist, but because explaining how you feel in a second language, to a stranger, in a system you don't understand, is its own kind of barrier. We built Unbound for students like Li: a 24-year-old MSc student from China at UCL, homesick, anxious about grades, struggling to sleep. She would never book a counsellor appointment — but she would speak to her phone in Mandarin for 90 seconds before bed.

What it does

Unbound is a voice-first mental wellbeing companion for international students. Users speak naturally in their own language, track their daily mood through a simple check-in, and get gently connected to professional support when patterns suggest they need it — with their emotional context already available in English for the therapist.

• Voice journal — speak in any language (Mandarin, Korean, Arabic, French, Spanish and more), transcribed in real time
• Daily mood check-in — 6 gentle questions mapping to clinical wellbeing indicators, no jargon
• 5-tier wellness system — communicated through colour and a cat mascot, never a clinical score or number
• Mood calendar — every day colour-coded, tap any date to view or edit
• AI pattern detection — quietly monitors for signs of sustained low mood, stress, or burnout
• Support page — warm, non-clinical resource cards connecting users to NHS Talking Therapies, Samaritans, Student Space, and more — with language tags surfacing resources in the user's native language
• Tier 5 safety card — if crisis signals are detected, professional resources auto-surface on the home screen. This cannot be hidden. It is a hard safety requirement we built in from day one.

Marketing video

Watch our Fotor marketing video

How we built it

We started with the problem, not the tech. The ideation and product structure was developed using Manus as a prototyping and thinking tool. The web app is built with React and TanStack Start, deployed on Cloudflare Workers with a Hono API backend and Cloudflare D1 for storage. Voice transcription uses Web Speech API for real-time multilingual input, with Deepgram API integrated as an upgrade path for higher accuracy. The design system uses FF Providence Sans and Montserrat with a warm cream palette — deliberately non-clinical.

Challenges

The hardest part wasn't the technology — it was the tone. Every word, colour, and interaction had to feel like a friend checking in, not a doctor assessing you. Making crisis detection feel safe rather than alarming took real care. Building a voice experience that works across 8 languages in a browser, without requiring an app install, was a meaningful technical challenge.

What we learned

Mental health support doesn't fail because the resources aren't there. It fails because the path to them is built for people who already know the language, the system, and how to advocate for themselves. Unbound is for everyone else.

Try the demo

Pre-seeded with 6 days of mood data so the dashboard, calendar, and AI summary are meaningful from the first load.

• Demo account: expel-scoop-exile@duck.com
• Password: mNL9FsYvE!3R

Built With

• canva
• claude
• cloudflare-d1
• cloudflare-workers
• drizzle-orm
• fotor
• hono
• manus
• react
• tailwind-css
• tanstack-start
• typescript
• vite
• web-speech-api