The Zeroic Journal

How to build a mental health app: features, privacy, cost

Mood tracking, journaling and a crisis path make a safe first version. Here is the build order, the privacy duties and where AI is off limits.

On this page 6 sections
  1. How to build a mental health app MVP
  2. Crisis paths are not optional
  3. Where AI is allowed and where it is banned
  4. Health data rules beyond HIPAA
  5. Scope by what your app promises
  6. How TopUp Care signed six clinics across four countries

How to build a mental health app: ship a 5-6 week MVP with mood check-ins, journaling, guided exercises and a crisis screen that reaches a human, then add clinicians later. Leave out AI that talks like a therapist. Illinois fines that up to $10,000, and Nevada up to $15,000 per violation.

The features are the easy part. What decides whether the app can launch is the safety path for users in distress, how you handle health data, and what you claim the app does. Get those three right in the first version and the rest is a standard product build.

How to build a mental health app MVP

Build the version that helps someone on a bad day without a clinician in the loop. That keeps scope small and keeps you clear of most regulation.

  • Accounts with a private default. Email sign-in, no social features, nothing visible to anyone but the user.
  • Mood check-ins. A daily prompt, a few taps, a history chart. This is the habit loop.
  • A private journal. Free text with optional prompts. It is also where your most sensitive data will sit.
  • Guided exercises. Breathing, grounding and short audio sessions. Scripted content is the safest thing you can ship, because it never improvises.
  • Reminders. Push or email, with wording that doesn’t reveal the app’s purpose on a lock screen.
  • A crisis screen. Covered in the next section. This is a feature, not a footer link.

Leave these for version two: booking with licensed therapists, video sessions, insurance billing, community forums and anything that analyzes what users write. Each one adds a regulated layer. Our telehealth app cost breakdown shows how those layers stack when you do add clinicians. A fixed-scope MVP Sprint covers the list above in 5-6 weeks.

Crisis paths are not optional

Some of your users will open the app in the worst hour of their year. Build for that on day one.

The US has a free, confidential answer you can point to. The 988 Suicide and Crisis Lifeline runs 24 hours a day, 7 days a week, by call, text and chat, through a network of more than 200 local crisis centers. Since September 17, 2024, cellular calls route to a local center based on the caller’s approximate location.

What that means for your build:

  • Reachable without an account. The crisis screen opens from the sign-in page as well as from inside the app.
  • One tap to call or text 988. Use native phone and SMS links. Don’t make the user copy a number.
  • A trigger inside the product. If a journal entry or message contains crisis wording, interrupt with the crisis screen. Keep the keyword list conservative and reviewed by a clinician.
  • Local numbers for each market. If you launch outside the US, replace 988 with that country’s service before the first user signs up.
  • No bot as the only response. Escalation always ends at a person or a hotline.

Apple’s App Review Guidelines add a store-level check. Guideline 1.4.1 says medical apps that could provide inaccurate information may get greater scrutiny, and that apps should remind users to check with a doctor before making medical decisions. Put that line in onboarding and in settings.

Where AI is allowed and where it is banned

Three states show the range, and the rules differ enough that your product claims decide your risk.

Illinois signed its Wellness and Oversight for Psychological Resources Act on August 1, 2025. As summarized by law firm Hepler Broom, it bars providing therapy or psychotherapy through internet-based AI, using AI to communicate with clients in a therapeutic way, and using AI to detect clients’ emotions or mental states. Administrative uses such as scheduling, reminders and billing are allowed, and so are supplementary tasks like record-keeping, with written informed consent when sessions are recorded or transcribed. Civil fines run up to $10,000, enforced by the state’s financial and professional regulator.

Nevada went further on marketing. Per Wilson Sonsini’s summary of AB 406, signed on June 5, 2025 and effective July 1, 2025, AI providers may not offer a system specifically programmed to deliver professional mental or behavioral health care. They also may not represent that the AI is a therapist, psychiatrist or other provider, and that includes any feature or avatar. Administrative tools for scheduling and billing are permitted. Civil penalties go up to $15,000 per violation.

Utah allows mental health chatbots but regulates them. Its H.B. 452 defines a covered chatbot as generative AI that holds conversations similar to those with a licensed therapist and that a reasonable person would believe can provide therapy or help manage mental health conditions. Scripted output such as guided meditations is excluded, and so is AI that only routes someone to a human therapist. Covered chatbots must say clearly that they are AI before the user can access features, again after seven days of inactivity and whenever the user asks. They can’t use a user’s input to target ads, and fines reach $2,500 per violation.

Other states and countries are writing their own versions, so check the current statutes for each market on the day you scope, and get legal advice. This section is not legal advice.

The build lesson is the same everywhere. Scripted content and journaling prompts carry the least risk. Use AI for admin work and for sorting users toward human help. Don’t ship an AI that diagnoses, treats or presents itself as a clinician.

Health data rules beyond HIPAA

Most wellness apps sold directly to consumers aren’t covered by HIPAA, which applies to providers, health plans and their vendors. That doesn’t leave the data unregulated.

The FTC’s Health Breach Notification Rule applies to vendors of personal health records, related entities and their service providers, and it excludes HIPAA-covered entities and their business associates. A consumer app that holds users’ health information can fall inside it. After a breach you must notify each affected person, and notify the FTC. Individual notice is due no later than 60 calendar days after you discover the breach, and media notice applies when 500 or more residents of a state are affected.

Three practical consequences:

  • Don’t send journal text or mood scores to ad or analytics tools. Apple’s guideline 5.1.3 bars using health data gathered in apps for advertising or data mining by third parties. Treat mood data the same way on every platform.
  • Collect less. Every field you store is a field you can leak. Skip date of birth, location and contacts unless a feature needs them.
  • Lock down the database. Row level security must be on, with policies that let each user read only their own entries. Our guides to Supabase row level security failures and Lovable app security issues list the settings to check.

If you do work with clinicians or health plans, HIPAA applies. Supabase’s HIPAA documentation requires a signed business associate agreement, says the hosted platform provides the needed controls while self-hosting doesn’t out of the box, and notes that Postgres connection logging is off by default for new projects and should be on for HIPAA projects. Every other vendor that touches patient data needs a signed agreement too.

Scope by what your app promises

Pick the row that matches your claim, then build only that scope.

  • “Track your mood and build a habit.” Check-ins, journal, reminders, crisis screen. No clinician, no AI conversation. Standard MVP.
  • “Guided exercises for stress and sleep.” Add scripted audio or text content. Still no AI conversation. Standard MVP.
  • “Find and book a therapist.” Add clinician profiles, booking, video and payments. The booking core is the same one behind an app like Calendly. You now handle clinical data, so plan for HIPAA, signed vendor agreements and licensing in each state.
  • “Support between therapy sessions.” Add clinician dashboards and consented sharing. Any AI stays on the admin side, such as summaries a clinician reviews.
  • “An AI companion that talks about your feelings.” Expect Utah-style disclosure rules and tight claims language. In Illinois and Nevada, calling it therapy or care is the risk. Take legal advice before you write a line of code, and gate by state if you must.

If you can’t say which row you’re in, you aren’t ready to write the spec. Decide the promise first.

How TopUp Care signed six clinics across four countries

We built TopUp Care, a telehealth platform for clinicians in Africa. It isn’t a mental health product, but it faced the same design problem: sensitive information, several kinds of users, and rules about who sees what. Doctors sent anonymous questions to a network of specialists, referred patients on and discussed cases in specialty chat rooms. It onboarded six hospitals and clinics across four countries, with over 100 doctors on the platform.

The platform worked because its rules lived in the product. Both identities stayed hidden in an eConsult so the answer was unbiased, providers and specialists got separate views, and every question needed at least 50 words of background. The same approach applies to a mental health app: write the privacy and safety rules first, then build screens around them. Our guide to fixed-price MVP development explains how we scope that first version so it ships on time.

Frequently asked questions

How much does it cost to build a mental health app?
A mental health MVP with sign-up, mood check-ins, journaling, guided exercises and a crisis screen is a fixed-scope build of about 5-6 weeks. At Zeroic that kind of build is an MVP Sprint, which starts at $12,000. Clinician booking, video visits, HIPAA-grade hosting and AI features are separate layers, so price them separately and add them after people use the first version.
Can I put an AI therapist in my mental health app?
Not safely in the US without legal advice. Illinois bars offering therapy through AI and fines violations up to $10,000. Nevada bars systems programmed to deliver professional mental health care and fines up to $15,000 per violation. Utah allows mental health chatbots but requires clear disclosure that the user is talking to AI, with fines up to $2,500 per violation.
Does a mental health app need to be HIPAA compliant?
Only if you handle health data for a covered entity such as a clinic, therapist or health plan, in which case every vendor needs a signed business associate agreement. A consumer wellness app sold directly to users usually sits outside HIPAA, but the FTC's Health Breach Notification Rule can still apply. It requires you to notify affected users within 60 calendar days of discovering a breach.
What does an app need to do when a user is in crisis?
It needs a visible path to human help that works without logging in. In the US that means showing the 988 Suicide and Crisis Lifeline, which takes calls, texts and chats around the clock. Detect crisis wording in journals and chat, interrupt with that screen, and never let a bot or an automated reply be the only response.
What features belong in a mental health app MVP?
Start with mood check-ins, a private journal, short guided exercises, reminders and a crisis screen. Those five cover most of what users open a wellness app for, and none of them requires a clinician or an AI model. Add booking with licensed therapists, video sessions and insurance billing only after people return to the first version weekly.

References

  1. Hepler Broom - Illinois passes legislation on using AI in delivering mental health services
  2. Wilson Sonsini - Nevada passes law limiting AI use for mental and behavioral healthcare
  3. Utah Legislature - H.B. 452 Artificial Intelligence Amendments (enrolled copy)
  4. Code of Federal Regulations - 16 CFR Part 318, Health Breach Notification Rule
  5. 988 Suicide and Crisis Lifeline - About 988
  6. Supabase - HIPAA compliance
  7. Apple - App Review Guidelines

More from the journal

Hear from Khushi within 24 hours.

Khushi Arora Partner @ Zeroic

Scoping a mental health app and want a fixed number? Our MVP Sprint ships check-ins, journaling, a crisis path and secure accounts to production in 5-6 weeks, with scope and price agreed upfront. Sprints start at $12,000. Bring your idea and we'll tell you what fits in version one.

See the MVP Sprint, or book a 30-minute call