Free. Open source. 13 languages. No sign-up. No tracking. No ads. Ever.
Use it now • What you get • The stack • Run it yourself • Join us • Sponsor
This project is free, open source, and runs on free-tier infrastructure. If MedOS is useful to you, sponsor on GitHub — every dollar helps keep healthcare guidance free for everyone, in every language.
Billions of people cannot afford a doctor. Millions search the internet for health answers and find ads, misinformation, or paywalls.
We believe the best medical AI in the world should be free, private, and available to every human on earth — regardless of where they live, what language they speak, or how much money they have.
This is that project. Built by the community. Free forever.
If that mission resonates, you can keep it going: sponsor on GitHub →
Important: This does not replace a doctor. It helps you understand your health, ask better questions, and know when to seek professional care.
The full medical AI experience — chat with the AI, track your health, manage medications, and more.
- Ask anything about your health
- 13 languages, auto-detected
- Voice input — speak your symptoms
- Emergency detection with local numbers
- Health tracker (meds, vitals, appointments)
- Medicine scanner (point your camera)
- Works on any phone as a PWA
- 100% private, zero data stored
Your personal health dashboard — private, encrypted, on your device.
- Medications with dose tracking and streaks
- Schedule timeline with live "now" indicator
- Vitals — blood pressure, glucose, temp, weight, heart rate, O2
- Appointments with doctor, location, and reminders
- Records — upload and organize health documents
- EHR Wizard — build your medical history in 5 steps
- Export everything as JSON or print for your doctor
You speak or type your symptoms
|
[ MedOS AI ]
|
Llama 3.3 70B ──── grounded in WHO, CDC, NHS guidelines
|
Clear answer in your language
|
Emergency detected? ──> local emergency number instantly
The AI runs on Llama 3.3 70B via free inference providers. It is grounded in medical guidelines from WHO, CDC, NHS, and major medical societies. It does not hallucinate treatment plans — it tells you what doctors would want you to know, and when to see one.
Everything here is free and open. Every piece can be run independently, improved, or adapted for your community.
Full medical platform — chat, health tracker, medicines, 13 languages
Fine-tuned Llama3 8B on 250K medical Q&As
AI that conducts a structured medical interview
AI trained to understand emotions (with Tilburg University)
IBM WatsonX medical symptoms analysis via MCP protocol
Point your camera at a medicine box — AI reads the label for you
| # | Project | What it does | Links |
|---|---|---|---|
| MedOS Web App | Full medical platform — chat, health tracker, medicine inventory, 13 languages |  |
|
| MedOS Backend | API server — auth, SQLite DB, LLM routing, emergency triage, RAG |  |
|
| 11 | Medicine Scanner | Camera scan of medicine labels via Qwen2.5-VL multimodal AI | |
| 12 | MetaEngine Nearby | Find nearby pharmacies and doctors via OpenStreetMap | |
| 13 | MedOS Family | Family health tree, adult consent, child mode, medicine reminders, family admin | |
| 14 | MedOS Connect | Additive backend for device + vitals sync (Withings, Apple Health, Health Connect), reminders, notifications | |
| 15 | MedOS Classify | MCP clinical-classification + triage engine — multi-head, calibrated, red-flag floor, Hugging Face models | |
| 16 | MedOS Pathogen | MCP pathogen-ID layer — microscopy / chest X-ray / genomic sequence with calibrated probabilities and visible evidence | |
| 17 | MedOS Research | R&D module — literature briefs, hypotheses, in-silico simulation plans, publication studio, safety reviewer | |
| Medical Llama3 8B | Fine-tuned Llama3 on 250K medical Q&A pairs |  |
|
| Medical Llama3 v2 | Improved medical Llama3 with enhanced chatbot interface | |
|
| Medical Mixtral 7B | Fine-tuned Mixtral for medical assistance | |
|
| Medical Interviewer | AI that conducts structured medical interviews and generates reports | |
|
| Empathy Chatbot | AI understanding emotions and unmet needs (Tilburg University) | |
|
| WatsonX MCP Server | IBM WatsonX medical symptom analysis via MCP protocol | |
|
| RAG Pipeline | Medical knowledge retrieval with FAISS / Milvus / ChromaDB | |
|
| Medical Dataset | 250K curated medical Q&A pairs for training | |
|
| Fine-tuning | Notebooks to fine-tune your own medical LLMs | |
The MedOS ecosystem is layered. The Core is what runs in production today — the web app, the backend API, the health tracker, the medicine scanner, and the nearby-care finder. The Ecosystem layer adds new modules on top, purely additively: each one ships as its own folder, behind a feature flag, with no destructive changes to existing files, tables, or APIs.
| Layer | Module | What it adds |
|---|---|---|
| Core | MedOS Web App | The medical platform users see today — chat, vitals, medicines, records, 13 languages. |
| Core | MedOS Backend | Next.js API + SQLite + LLM routing + medical RAG + emergency triage. |
| Core | Health Tracker | Vitals, medicines, appointments, records, EHR wizard, JSON / print export. |
| Core | Medicine Scanner | Camera → structured JSON via Qwen2.5-VL. |
| Core | Nearby Care | Pharmacies and doctors via OpenStreetMap with ranking and route metadata. |
| Ecosystem · 13 | MedOS Family | Family health tree, adult consent, child mode, medicine reminders, family admin. |
| Ecosystem · 14 | MedOS Connect | Device + vitals sync (Withings, Apple Health, Health Connect), reminders, notifications. |
| Ecosystem · 15 | MedOS Classify | MCP triage tools — multi-head, calibrated, red-flag-gated; ships through Hugging Face. |
| Ecosystem · 16 | MedOS Pathogen | MCP pathogen ID — microscopy, chest X-ray, genomic sequence; visible evidence. |
| Ecosystem · 17 | MedOS Research | R&D module — literature briefs, hypotheses, in-silico simulation plans, safety-reviewed publication studio. |
Every Ecosystem module follows the same contract: additive only, namespaced, feature-flagged, reversible, never auto-flowing patient PHI across boundaries.
| Vercel | HuggingFace Space | |
|---|---|---|
| Role | Frontend only | Frontend + Backend + DB |
| Cost | Free | Free |
| Deploy | git push |
bash scripts/deploy-hf.sh |
Just click. No account. No install. Free.
git clone https://github.com/ruslanmv/ai-medical-chatbot.git
cd ai-medical-chatbot/web
npm install && npm run devOpen http://localhost:3000. Done.
# Deploy to HuggingFace Spaces (free)
HF_TOKEN=hf_xxx bash 9-HuggingFace-Global/scripts/deploy-hf.sh
# Deploy to Vercel (free tier)
# Root Directory = web, Framework = Next.js
# Env: NEXT_PUBLIC_BACKEND_URL = your HF Space URLcd 6-FineTunning
# Follow the notebooks to fine-tune Llama3 or Mixtral
# on 250K medical Q&A pairsThis project exists because people like you contribute. Every fix, every translation, every idea makes healthcare more accessible for someone, somewhere.
Ways to help:
- Add a language translation (we support 20, the world has 7,000)
- Improve medical knowledge accuracy
- Report bugs or suggest features
- Share the project with someone who needs it
- Star the repo so others find it
git clone https://github.com/ruslanmv/ai-medical-chatbot.git
cd ai-medical-chatbot
git checkout -b my-improvement
# Make your changes, then open a Pull Request
MedOS is free for users forever. It runs on free-tier cloud and gets better only because people contribute time, translations, clinical review, and — increasingly — sponsorship to keep the lights on.
If MedOS has helped you understand a symptom, prepare for a doctor's visit, or just felt like a calm voice on a rough day, please consider supporting it.
Both help. Sponsorship pays for hosting beyond the free tiers, the LLM provider keys that keep the chat fast in every language, and the clinician review hours that keep the safety floor honest. Stars help the next person who needs MedOS actually find it.
You can also help without spending anything:
- Translate one screen into a language we don't fully cover.
- Add one safety test case to
tests/safety/golden_prompts.jsonl. - Tell a friend who could use this.
Thank you. Every form of support, including a star, keeps this open.
Apache 2.0 — use it, modify it, deploy it, share it. Free forever.
Made with care by Ruslan Magana Vsevolodovna and the open source community
The best technology in the world should serve everyone — not just those who can pay for it.
Let us use it to heal.
