MindCore
A nonprofit foundation building simple, safe AI tools for mental wellbeing—so people get structured support in daily life, not slides and theory.
Product-first · Safety layers · Real user feedback
(An early version was shared at the AI GENIUS Olympiad.)
Open source on GitHub
Browse the MindCore codebase, report issues, or contribute.
The Problem
Mental health remains one of the most underserved areas in global healthcare
Over one billion people worldwide lack access to any form of structured support. The gap between those who need help and those who receive it continues to widen.
1 in 3
Rising anxiety among students
college students report experiencing significant anxiety, with rates doubling over the past decade.
60%
Limited access to therapists
of young people who need mental health support do not receive any treatment due to cost, stigma, or availability.
< 5%
Lack of daily emotional monitoring
of mental health apps provide science-backed daily emotional monitoring with actionable insights.
The Solution
MindCore: daily support you can actually use
A focused product for mood tracking, guided reflection, and AI-assisted insight—with efficient models, clear routing, and crisis-aware safeguards so the experience stays practical and responsible.
Self Awareness
Know yourself better, one check-in at a time
MindCore helps users build emotional vocabulary and recognize patterns through daily mood tracking, guided emotion labeling, and reflective check-ins. The more you reflect, the more clearly you see your inner landscape.
Guided Reflection
Structured prompts grounded in CBT practice
Built on Cognitive Behavioral Therapy frameworks, MindCore offers guided reflection sessions that help users challenge negative thoughts, identify cognitive distortions, and develop healthier thinking patterns over time.
AI-assisted Insights
Your data, interpreted with intelligence and care
MindCore's AI engine analyzes reflective writing and mood data to surface meaningful patterns, emotional triggers, and behavioral trends. These insights empower users to take informed steps toward emotional resilience.
How It Works
Four steps to better mental wellness
Track your emotions
Log how you feel throughout the day with quick, intuitive mood check-ins designed to build self-awareness.
Reflect through guided prompts
Respond to CBT-based prompts that help you explore and understand your emotional triggers and thought patterns.
Identify mental patterns
The AI engine analyzes entries over time to surface recurring emotional trends and behavioral insights.
Improve emotional regulation
Receive personalized strategies and exercises to better manage stress, anxiety, and emotional well-being.
Technology
Technology behind MindCore
Built with modern tools and a focus on privacy, scalability, and real-world usability. Select a layer to explore.
Cross-Platform Access
Users can access MindCore from any device via a responsive web app optimized for mobile, tablet, and desktop.
Privacy-First Design
All user data is anonymized and encrypted. Users maintain full control over their data with easy export and deletion options.
Intuitive Experience
A warm, minimal interface designed to feel safe and inviting. No clinical jargon -- just clear, human-centered language.
Personal Dashboard
Users get a personalized dashboard showing mood trends, reflection streaks, wellness scores, and AI-generated weekly summaries.
AI Architecture
How the fine-tuned models work
MindCore uses supervised fine-tuning on large language models to create an AI that responds with empathy, applies CBT techniques, and prioritizes user safety.
Data Collection
The training corpus contains thousands of anonymized therapy transcripts and mental health Q&A pairs, each labeled with emotional tone, CBT category, and response quality ratings.
Data Formatting
Conversations are structured into system/user/assistant message triplets following the ChatML format. Each example includes context about the user's emotional state and the therapeutic approach.
Fine-Tuning Process
Supervised fine-tuning (SFT) on a base language model yields a system that responds empathetically, applies CBT techniques, and recognizes crisis signals, with multiple training epochs and careful validation.
Evaluation & Safety
Each model version is evaluated against safety benchmarks, empathy scoring rubrics, and CBT adherence metrics. Red-team testing ensures the model handles sensitive topics responsibly.
Training Data Sample
mindcore-training-data-sample.json
{
"messages": [
{
"role": "system",
"content": "You are a supportive mental wellness assistant using CBT techniques, be empathic, calm, and non-judgmental. Acknowledge emotions then ask a gentle question. Do not assume names, or context."
},
{
"role": "user",
"content": "I feel responsible for keeping the peace all the time and it's exhausting."
},
{
"role": "assistant",
"content": "Carrying that responsibility all the time sounds genuinely exhausting. What do you think would happen to the relationship if you stopped?"
}
],
"metadata": {
"cbt_technique": "exploratory_questioning",
"emotion_labels": ["exhaustion", "responsibility", "people_pleasing"],
"severity": "moderate"
}
}This sample shows the ChatML format used for training. Each example includes metadata for CBT technique classification, emotion labeling, and severity assessment.
Therapeutic Framework
Cognitive Behavioral Therapy explained
CBT is one of the most evidence-based approaches to mental health. It focuses on the connection between thoughts, feelings, and behaviors -- and MindCore brings it to your fingertips.
The CBT Cycle
CBT works by breaking the cycle of negative automatic thoughts that reinforce harmful emotions and behaviors.
CBT intervenes at the Thought stage to break negative cycles
Identify the Thought
The first step in CBT is recognizing automatic negative thoughts (ANTs) -- the quick, reflexive thoughts that arise in response to stressful situations. MindCore's reflective prompts are designed to help users catch these thoughts as they happen.
"I messed up the presentation, everyone thinks I'm stupid."
Examine the Evidence
Once a thought is identified, CBT encourages examining the actual evidence for and against it. MindCore guides users through structured questions that challenge the validity of distorted thinking.
"What evidence do I have? Did anyone actually say I was stupid? Two people complimented my slides afterward."
Cognitive Restructuring
The core of CBT: replacing distorted thoughts with more balanced, realistic alternatives. MindCore's AI suggests reframes based on the specific cognitive distortion detected in the user's written reflection.
"I felt nervous and stumbled on one point, but overall I delivered the content. Nervousness is normal."
Thought Recording
CBT uses structured thought records to track the cycle of situations, thoughts, emotions, and behaviors. MindCore digitizes this process, making it easy to maintain and review over time.
Situation -> Automatic thought -> Emotion -> Evidence -> Balanced thought
Behavioral Experiments
CBT encourages testing beliefs through real-world experiments. MindCore suggests small, manageable actions users can take to challenge their assumptions and build confidence through direct experience.
"Next time, I'll ask one person for honest feedback right after presenting."
Common Cognitive Distortions
MindCore's AI is trained to identify these patterns in user reflections and gently help users recognize them.
All-or-Nothing Thinking
Seeing things in black and white: "If I'm not perfect, I'm a total failure."
Mind Reading
Assuming you know what others are thinking without evidence.
Catastrophizing
Expecting the worst possible outcome in every situation.
Overgeneralization
Using a single event to form a sweeping conclusion: "This always happens."
Emotional Reasoning
Believing something is true because it feels true: "I feel dumb, so I must be dumb."
Should Statements
Rigid rules about how things must be: "I should never make mistakes."
Safety Protocol
How MindCore handles crisis situations
User safety is the highest priority. MindCore is not a replacement for professional help, and its crisis detection system is designed to recognize distress signals and direct users to real support immediately.
Real-Time Detection
Every journal entry and mood check-in is analyzed in real-time by an NLP pipeline for crisis indicators. The model is trained to detect language patterns associated with self-harm ideation, suicidal thoughts, severe distress, and abuse disclosures.
Keywords, sentiment extremes, and contextual cues trigger the detection system. A multi-layer scoring approach minimizes false positives while avoiding missed distress signals.
Risk Stratification
Detected signals are scored across three severity levels: Low (elevated distress), Medium (expressed hopelessness or withdrawal), and High (direct crisis language). Each level triggers a different intervention pathway.
Low: supportive prompts and grounding exercises. Medium: safety plan presentation and resource links. High: immediate crisis resource display with direct hotline access.
Immediate Intervention
When a High severity signal is detected, MindCore immediately pauses normal interaction and presents a dedicated crisis support screen. The AI does not attempt to provide therapy in crisis situations.
The crisis screen includes one-tap access to local crisis hotlines, text-based crisis services, a guided breathing exercise, and a clear message that professional help is available and recommended.
Follow-Up Protocol
After a crisis event, MindCore adjusts its approach for subsequent sessions. Check-in prompts become gentler, reflective suggestions focus on safety and stability, and the app periodically re-presents crisis resources.
The system maintains a cooldown period where it avoids deep emotional exploration and instead focuses on grounding, routine, and connection to real-world support.
Safety Principles
Never replace professional help. MindCore always defers to licensed professionals for crisis intervention.
No AI therapy in crisis. When distress is detected, the AI stops generating therapeutic responses and shows direct crisis resources.
Err on the side of caution. The detection system is calibrated to be sensitive rather than specific, ensuring no genuine distress is overlooked.
Transparent communication. Users are clearly informed about the limitations of AI support during onboarding and throughout the app.
Why we exist
Why this matters
Millions of people still lack day-to-day mental health support. MindCore exists to put thoughtful, product-grade AI in their hands—so everyday emotional care is easier to access without replacing clinicians or crisis services.
Mental Health Accessibility
Bridging the gap between those who need support and those who can access it, regardless of location or financial status.
Digital Early Detection
Leveraging AI to identify emotional patterns that may indicate developing mental health concerns before they escalate.
Scalable Support
Providing a cost-effective complement to traditional therapy that can scale to millions of users worldwide.
App Preview
See MindCore in action
Explore the core interfaces that make MindCore a powerful companion for emotional well-being.
Coming Soon
MindCore for Mobile
The full MindCore experience is coming to iOS and Android. Track your mood, capture reflections, and receive AI insights -- all from your pocket.
Native Experience
Built natively for iOS and Android with smooth animations, haptic feedback, and platform-specific design patterns.
Smart Reminders
Gentle push notifications to check in with your emotions at the right moments, adapting to your schedule and habits.
Offline Support
Capture reflections and track your mood even without an internet connection. Data syncs automatically when you reconnect.
Mission
Practical AI for mental health, built to ship
MindCore is a nonprofit foundation focused on launching usable mental health tools—not positioning as a research lab or generic community hub. We combine modern language models with structured safety layers and clear, user-centered design so support stays accessible, transparent, and grounded in real feedback as we iterate. An early version was also presented at the AI GENIUS Olympiad.
Evidence-informed product
CBT and mental-health best practices inform what we build—so guidance stays structured and responsible, not experimental copy on its own.
Safety by design
Layers for routing, crisis awareness, and data handling are part of the product, not an afterthought—so the experience stays safer as we scale.
Execution over theory
We prioritize shipping, learning from real use, and improving the tool in the open rather than staying in slide-deck mode.
Watch the presentation from the AI GENIUS Olympiad
FAQ
Frequently asked questions
Everything you need to know about MindCore.
MindCore is an AI-assisted mental health companion designed for young people. It combines mood tracking, guided reflection based on Cognitive Behavioral Therapy (CBT), and AI-powered emotional insights to help users build emotional awareness and resilience.
No. MindCore is a wellness tool, not a medical device or therapy replacement. It is designed to complement professional mental health support by providing daily self-reflection tools and emotional awareness exercises. If you're in crisis, please reach out to a licensed professional or crisis hotline.
MindCore uses Natural Language Processing (NLP) to identify emotional tone, sentiment patterns, and potential cognitive distortions in your writing. The AI surfaces trends and patterns over time, helping you see your emotional landscape more clearly without making diagnoses.
Absolutely. Written reflections and mood data are encrypted using AES-256 at rest and TLS 1.3 in transit. Your data is never shared with third parties. You maintain full control with easy export and deletion options at any time.
MindCore is a nonprofit foundation focused on shipping practical AI products for mental health: accessible tools with safety layers and honest UX. An early version was also presented at the AI GENIUS Olympiad, but the through-line is real-world product impact—not competition framing.
MindCore is built with Next.js and React on the frontend, a RESTful API layer with serverless functions, PostgreSQL for data storage, and custom NLP models for emotional analysis. The entire stack is designed for privacy, scalability, and cost efficiency.
Yes. Today the app is offered as a free demonstration while we iterate. Efficient infrastructure keeps running costs low so we can scale access responsibly as the product matures.