CalmSpace

Inspiration

We've all been there—burning the midnight oil, skipping breaks, sleeping poorly, and telling ourselves, "I'll fix my habits tomorrow." Tomorrow rarely comes.

We built CalmSpace because we noticed that most wellness apps demand too much: accounts, lengthy onboarding, detailed personal data, and constant notifications. The barrier to starting is exactly what stops people from starting.

The UN Sustainable Development Goal 3 (Good Health and Well-Being) inspired us to ask:

What's the smallest possible action that could still make a real difference?

The answer was a single daily check-in—no account, no pressure, just one honest minute with yourself.


What it does

CalmSpace is an anonymous AI-powered mental wellness companion. Users can:

  • 🌤️ Check in daily by selecting their current mood (no account required)
  • 🤖 Talk to an AI companion powered by Claude—empathetic, non-judgmental, and available 24/7
  • 🧘 Practice breathing & grounding exercises like:
    • 4-7-8 Breathing
    • 5-4-3-2-1 Grounding Technique
  • 📖 Journal privately during the session—entries stay local in the browser and are never sent to any server
  • 🆘 Access crisis resources with global helplines for immediate support

Privacy First

  • ✅ Zero data collection
  • ✅ Zero tracking
  • ✅ Fully anonymous

How we built it

We used a modern, single-repo stack designed for speed, maintainability, and clarity.

Tech Stack

  • Frontend & Backend: Next.js 14 (App Router + API Routes)
  • Styling: Tailwind CSS
  • AI Engine: Anthropic Claude (claude-sonnet-4-6)
  • Storage: sessionStorage only
  • Deployment: Vercel

The Claude system prompt was carefully engineered to be:

  • Empathetic
  • Supportive
  • Non-judgmental
  • Crisis-aware
  • Clear about not replacing professional care

The project structure was intentionally modular, separating:

  • UI Components
  • API Routes
  • Utility Functions
  • Shared Logic

This also allowed PresentMe's AI analyzer to understand and document the project more effectively.


Challenges we ran into

Prompt Engineering for Safety

Our biggest challenge was making Claude warm and supportive without crossing ethical boundaries.

It required many prompt iterations to ensure the AI would:

  • Never diagnose
  • Never prescribe treatment
  • Never claim to replace therapists
  • Encourage professional help when appropriate
  • Escalate to crisis resources when distress signals appear

Privacy by Design

Every technical decision had to answer one question:

"Does this store anything about the user?"

If the answer was yes, we redesigned it.

Choosing no databases, no analytics, and no user accounts felt unconventional—but for a mental health application, it was the right decision.


Scope Creep

Like every hackathon project, ideas kept growing.

Some features we intentionally postponed include:

  • Mood trend analytics
  • Multilingual support
  • Push reminders
  • Community features

Instead, we stayed focused on one principle:

Simple. Anonymous. Helpful.


Accomplishments that we're proud of

  • 🎉 Built a genuinely useful mental wellness tool with zero data collection
  • 🤖 Designed an AI companion that feels supportive without pretending to be a therapist
  • 🧩 Created a clean, modular codebase that's easy to understand and maintain
  • 🚀 Delivered a complete experience during the hackathon:
    • Mood check-ins
    • AI conversations
    • Guided exercises
    • Private journaling
    • Crisis resources
  • ❤️ Maintained strong ethical values throughout:
    • Open Source (MIT)
    • No tracking
    • No ads
    • Privacy first

What we learned

Prompt Design is Product Design

The system prompt shaped nearly every part of the experience.

Tone, empathy, safety, boundaries, and helpfulness all came from thoughtful prompt engineering.


Constraints Breed Creativity

Working without:

  • Databases
  • Accounts
  • Analytics

forced us to focus only on what truly benefits someone during a vulnerable moment.


Mental Health UX is Different

Small design choices matter.

Calming colours, gentle language, reduced friction, and respectful interactions aren't optional—they're essential to making users feel safe enough to stay.


What's next for CalmSpace

We're excited to continue improving CalmSpace beyond the hackathon.

🌍 Multilingual Support

  • Interface in 10+ languages
  • Localized crisis resources

📊 Anonymous Mood Insights

Optional mood trends that preserve complete anonymity without identifying users.

🤖 Proactive AI Nudges

Opt-in browser notifications that gently encourage daily check-ins.

🧘 Expanded Wellness Library

Additional guided exercises including:

  • Body Scan Meditation
  • Progressive Muscle Relaxation
  • Sleep Meditations
  • Mindfulness Sessions

🤝 Community Partnerships

Integrating verified crisis helpline APIs to provide real-time availability and trusted support resources.


Closing Thoughts

CalmSpace started as a hackathon project.

It ended as something we'd genuinely use ourselves.

A simple, private, and compassionate space where taking care of your mental well-being starts with just one honest minute.

Built With

Share this project:

Updates