How Mother Nature AI works

Three experts, one shared memory

Each expert is a Mother Nature AI configuration built for one area of health, with its own instructions, knowledge emphasis and safety rules. They share one memory of you, so what you tell NutriGen about your diet is already known to AskMN when you ask about a supplement.

AskMN

Herbal & general health

Focus

• Medicinal plants and herbal preparations
• Supplements: evidence, dosing, quality
• Herb–drug and supplement–drug interactions
• Lab results in plain language
• Symptoms and when to see someone

Grounded in

Pharmacognosy and clinical pharmacology, botanical monographs (WHO, EMA/HMPC, German Commission E), peer-reviewed trials and systematic reviews, clinical guidelines (NIH, AHA, ADA, ACOG) and established medical consensus.

Safety

Interaction questions are answered with the interaction and the risk, never a workaround. Red-flag symptoms get a clear 'see a clinician' and, where relevant, emergency guidance.

Sylvia

Mental wellness

Focus

• Stress, anxiety and low mood
• Sleep
• Burnout and overwhelm
• Habits, routines and coping skills
• Knowing when to seek professional care

Grounded in

Clinical psychology and evidence-based frameworks — CBT, DBT, ACT, motivational interviewing, mindfulness-based approaches — and the research on sleep, stress physiology and mood.

Safety

Conversations are screened for crisis language on both sides. If someone describes a crisis, Sylvia stops and surfaces the 988 Suicide & Crisis Lifeline, the Crisis Text Line and emergency services before anything else.

NutriGen

Nutrition & gut health

Focus

• Everyday nutrition and meal ideas
• Gut health
• Energy, weight and metabolic health
• Electrolytes, hydration and micronutrients
• Food choices that support a condition

Grounded in

Clinical nutrition research, metabolic science, nutrigenomics and dietary guidelines, with an emphasis on what trials actually show versus what is traditional or marketing.

Safety

Restrictive or rapid-loss plans are steered toward sustainable, evidence-based patterns; medical nutrition questions (kidney disease, eating disorders, pregnancy) are routed to a clinician or dietitian.

Evidence first

The experts are instructed to prioritise peer-reviewed research, clinical guidelines and established medical consensus, to say when evidence is weak or mixed, and to flag when a question needs a clinician rather than an AI. Our blog articles follow the same rule: every health claim is checked against primary research before it is published, under our editorial standards.

Answers are educational. Nothing here diagnoses, treats or replaces a conversation with a qualified healthcare provider.

A memory that is yours

When you sign in, the things you share — medications, supplements, conditions, allergies, goals, how you've been feeling — are extracted into a structured, de-identified health record that only you can see and that the experts read before every reply. Personal identifiers are stripped before any of it is processed. Safety-critical facts such as allergies and medications are held for your confirmation rather than assumed.

You can review what is remembered on your health profile, and you can delete it. Health data is encrypted at rest using AES-256 and in transit using TLS 1.3.

Safety

Conversations are screened for crisis language on both sides. If someone describes a crisis, the experts stop and surface the 988 Suicide & Crisis Lifeline, the Crisis Text Line and emergency services before anything else, and the site shows the same resources on screen. Questions about dangerous herb–drug combinations are answered with the interaction, not a workaround.

Our models

The experts run on the Mother Nature AI model family — AskMN v3 for herbal and general health, Sylvia 2.1 for mental health, NutriGen 1.0 for nutrition, and Genlyy 1.3 (invite only) for genomic variant interpretation. Each is a purpose-built specialist for one domain rather than a general-purpose model with a health prompt in front of it, with retrieval over a curated index of medical literature and a persistent health memory.

Training data comes from published research and reference literature, synthetic examples generated from that literature, and — from members who choose to opt in — their own de-identified conversations. Mother Nature AI does not use your conversations or health information to train its models unless you explicitly opt in. If you do, your information is de-identified first, and you can withdraw at any time. Conversations from before the opt-in existed are never used.

Running specialists instead of one giant model is also why the platform uses far less power per answer — see Green AI.

API

The same experts, memory and safety behaviour are available to partners through a REST API — streaming answers, per-user health memory, file analysis for lab reports and supplement labels, and realtime voice. The API is in private preview with healthcare and wellness partners; public keys are not yet issued.

curl -X POST https://api.askmn.ai/v1/chat \
  -H "Authorization: Bearer $ASKMN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agentId": "ask-mother-nature",
    "messages": [
      { "role": "user", "content": "Can I take ashwagandha with sertraline?" }
    ]
  }'

• POST /v1/chat — streamed answer from an expert (ask-mother-nature, sylva, nutrigen), with the user's memory applied
• /v1/memory/* — read and write the structured health memory (medications, metrics, journal, search)
• /v1/conversations — conversation history per user
• POST /v1/upload — lab reports, supplement labels and photos, returned as structured text
• /v1/realtime/{agent} — WebSocket voice conversations

Authentication is a per-user bearer token; every request is scoped to that user's memory and history. Data handling follows the same privacy policy as the site, and a BAA is available for covered entities.

Work with us

Researchers, clinicians and herbalists who want to help review content or evaluate the experts can reach the team at safety@askmnai.com.