# getbased

> Free, open-source, local-first Personal Health Intelligence that runs in your browser.
> getbased connects five lenses on your biology — labs, genome, body, light, and
> insight — and lets AI reason across the whole picture, grounded in the research
> you trust. Import any lab PDF or photo with AI, factor in your DNA, sync your
> wearables, log your sun exposure. No account required. Core health data stays on
> your device by default; optional network features send only the data needed for
> the feature you deliberately use.

## The five lenses

getbased is organized around five lenses on your biology. Each one informs the others, and the AI synthesizes across all of them — a gene variant reframes a lab result, your light exposure explains a body signal, and Biology Scores summarize current patterns.

- **Labs** — blood work, OAT, fatty acids, BioStarks and other specialty panels. 287+ markers across 17+ categories, reference and optimal range bands, trend detection
- **Genome** — raw DNA + mtDNA import from any major provider. 50+ curated SNPs across 13 health categories with APOE haplotype, plus mtDNA haplogroup with 39 sub-clades and mitochondrial coupling classification
- **Body** — wearables (Oura, Fitbit, Withings, Polar, Apple Health) plus manual weight, blood pressure, and pulse. HRV, sleep, resting heart rate, body composition, vascular age, SpO₂, and an optional Meals & Nutrition log
- **Light** — Light & Sun exposure tracking across six biological channels, plus on-device measurement tools
- **Insight** — AI chat, research library, correlations, and Biology Scores

## What getbased does

- Import any blood work PDF, photo, or screenshot — AI extracts all markers automatically (any format, any language, batch import). Specialty labs supported: OAT (organic acids), fatty acids (Spadia, ZinZino, OmegaQuant), BioStarks
- Track trends over time with interactive charts across 17+ biomarker categories (287+ markers), with reference + optimal range bands and automatic trend flags
- Multi-unit display — see every value in both mg/dL and mmol/L (and mU/L vs µIU/mL), per-profile preference
- Per-value notes — attach context like "fasted 14h" to any individual reading; the AI sees notes grouped by marker
- DNA import — raw data from any major provider, including MyHeritage Low-pass WGS with strand-aware SNP matching. 50+ curated SNPs across 13 health categories with APOE haplotype resolution
- mtDNA haplogroup import — 39 haplogroups with sub-clade resolution, mitochondrial coupling classification (Doug Wallace's framework), and environment mismatch detection. Manual haplogroup entry also supported
- Light & Sun lens — track sun exposure across six biological channels (vitamin D, body clock, cardiovascular, mood & hormones, cellular repair, outdoor eye light). One-tap session logging, live burn-dose estimates, photosensitizing-medication awareness, and eight on-device measurement tools (lux meter, flicker detector, color-temperature, sleep darkness meter, sunrise/sunset logger, eye-level audit). Camera frames never leave the device
- Meals & Nutrition — optional photo-assisted meal log with rolling 7/30-day nutrition aggregates. Two capture modes: meal photo (AI estimates components and nutrients, user reviews and corrects before save) and nutrition label (printed values scaled to the amount eaten). Estimates are always editable; unknown nutrients are omitted, not zeroed. Requires an AI provider for photo/label analysis; manual entry works without one
- Wearable integrations — the hosted app supports Oura, Withings, Polar, existing legacy Fitbit connections, and local file imports. Supported hosted OAuth/API calls use a fixed compatibility service on Company-operated Czech VPS infrastructure after a provider-specific explicit-consent screen; WHOOP, Ultrahuman, and Google Health remain self-host only. HRV, sleep, resting heart rate, body composition, vascular age, blood pressure, and SpO₂ appear alongside lab data. Every wearable token is device-encrypted. Raw daily samples stay on-device for ordinary app use; only a compact end-to-end-encrypted summary syncs (~200 tokens per profile). Users can separately opt an MCP agent into a 7-, 30-, or 90-day pivoted daily series for time-series reasoning
- Medical & Family History — 117 conditions across neuro, cancer, skin, mental health, autoimmune and more, plus first-degree relatives and grandparents (with onset age and notes) so the AI sees what runs in the family
- Biology Scores and Biological Coherence — deterministic pattern summaries across metabolism, thyroid, inflammation, methylation, iron, hormones, kidney/hydration, liver/bile, cellular energy, and more
- Biological age — PhenoAge (Levine 2018) + Bortz Age combined, with chronological age reference line, plus auto-computed derived markers and ratios
- Biometrics tracking — weight, blood pressure, pulse with time-series history and auto-calculated BMI
- Baubiologie EMF assessment — room-by-room measurements graded against SBM-2015 building-biology standards
- Bring-your-own research library — drop PDFs, research papers, or clinical guides into the app and the AI grounds every answer in those sources instead of relying only on its training. Runs fully on-device (browser RAG via MiniLM + OPFS), with multi-library support and query rewriting for better recall. Optional external-server backend for larger corpora
- AI chat panel with debate mode — pick an analytical lens, generate a persona’s style, and pit two against each other. Conversation threads, image attachments, round-by-round steering, web search hints. Always an AI simulation, never a real clinician
- 9 lifestyle context cards (diet, sleep, exercise, stress, love life, environment, light, medical & family history, goals) shape AI interpretation
- DNA-aware and region-aware supplement and lifestyle education (context to discuss with a healthcare professional, not dosing advice)
- Multi-profile support for practitioners, families, or multiple clients
- Menstrual cycle tracking with phase-aware reference ranges, perimenopause pattern insights, iron trend flags
- Correlation viewer, heatmaps, compare dates, trend flags, PDF reports
- E2E encrypted cross-device sync via Evolu CRDT with sync status indicator — self-host the relay for full control
- Personal agent integration — query your labs from Hermes Agent, OpenClaw, or any MCP-compatible bot over Telegram, Signal, Matrix, or any messenger. Access is read-only and user-enabled; a compact summary is standard, with a separate optional 7-, 30-, or 90-day wearable series
- Local-first — core data is stored in the browser; optional passphrase protection uses AES-256-GCM for sensitive fields, every wearable token and Google Health row is device-encrypted, and encrypted backup recovery is available
- Available over Tor as a hidden service, with Onion-Location discovery
- AI providers: PPQ, Routstr, OpenRouter, Venice, Local AI (Ollama, LM Studio, Unsloth Studio, Jan), or a custom OpenAI-compatible endpoint. PPQ, Routstr, and Venice offer optional verified E2EE (Private TEE / encrypted TEE modes on supported endpoints). Cloud AI requires provider-specific approval before the first data-bearing request

## Key pages

- [Landing page](https://getbased.health/): Feature overview and getting started
- [Web app](https://app.getbased.health): Launch the dashboard (no install needed)
- [Documentation](https://docs.getbased.health): User guides, setup, and feature walkthroughs
- [Blog](https://getbased.health/blog): Development stories and health data insights
- [GitHub](https://github.com/elkimek/get-based): Source code (AGPL-3.0 license)
- [Discord](https://discord.gg/zJdVB9zgQB): Community and support

## Docs highlights

- [Getting started](https://docs.getbased.health/quickstart): First import walkthrough
- [AI setup guide](https://docs.getbased.health/guides/ai-chat): Configure PPQ, Routstr, OpenRouter, Venice, or Local AI
- [Interpretive Lens (research library)](https://docs.getbased.health/guides/interpretive-lens): Add your own PDFs, papers, and guides so the AI grounds answers in sources you trust
- [DNA import](https://docs.getbased.health/guides/dna-import): Upload raw DNA data, mtDNA haplogroup, and view genetic insights
- [Cross-device sync](https://docs.getbased.health/guides/cross-device-sync): E2E encrypted sync setup, status indicator, and self-hosting
- [Agent access](https://docs.getbased.health/guides/agent-access): Connect Hermes, OpenClaw, or any MCP-compatible agent over any messenger
- [Wearables](https://docs.getbased.health/guides/wearables): Browser-direct and file-import options on the hosted app, plus relay-dependent providers on self-hosted deployments
- [Privacy and security](https://docs.getbased.health/guides/encryption): Local-first architecture, encryption, PII obfuscation

## What getbased is not

- **Not a medical device** — getbased is health intelligence software, not a diagnostic tool or telemedicine service. It does not diagnose, treat, or prescribe. Always consult a qualified healthcare provider for medical decisions.
- **Not a substitute for professional medical advice** — AI interpretations are structured reasoning aids grounded in your data and your research library, not medical conclusions. The human is always in the loop.
- **Not a central health-data account** — core data is stored locally in your browser by default. No account or central plaintext health database. Optional cloud AI, voice, wearables, encrypted sync/sharing, and other network features run only when you choose them.
- **Not a supplement store** — getbased provides DNA-aware supplement and lifestyle education (context to discuss with a healthcare professional, not dosing advice), but does not sell or affiliate-market products.
- **Not a genetic testing lab** — DNA analysis requires importing raw data from an external provider (23andMe, Ancestry, MyHeritage, etc.). getbased does not sequence DNA.

## Common misconceptions

- **Misconception**: "getbased is a chatbot that gives medical advice."
  **Correction**: It is a structured data harness with AI reasoning constrained by your actual lab values, reference ranges, and research library. The AI explains patterns; it does not diagnose.
- **Misconception**: "My data is uploaded to a server."
  **Correction**: Core data stays in your browser (localStorage/IndexedDB/OPFS) by default. Optional network features are explicit. Cross-device sync sends end-to-end-encrypted ciphertext that the relay cannot read.
- **Misconception**: "getbased sends my health data to OpenAI."
  **Correction**: You choose your AI provider (PPQ, Routstr, OpenRouter, Venice, Local AI, or custom endpoint). Cloud requests can include the prompt, enabled profile context, and attachments you choose; the app asks for provider-specific approval before the first request. Local AI does not require cloud approval.

## How getbased differs from alternatives

- **vs InsideTracker / SelfDecode**: getbased is free, open-source, and local-first. No subscription or account is required; optional network features are disclosed, the public landing page asks before loading cookieless analytics, and hosted-app analytics have an immediate persistent opt-out
- **vs spreadsheets**: AI-powered PDF import extracts markers automatically. Interactive charts with reference/optimal bands, trend flags, and correlation analysis
- **vs generic health trackers**: getbased connects five lenses on one dashboard — labs (287+ biomarkers, specialty panels), genome (DNA + mtDNA), body (wearables + biometrics), light, and insight — and the AI reasons across all of them, not one signal at a time
- **vs ChatGPT/Claude directly**: getbased structures your data, maintains history across imports, gives AI your full health context, and grounds every answer in your own research library (PDFs, papers, clinical guides) — not just a single lab report pasted into a chat window or what a generic model was trained on. Connect via Hermes, OpenClaw, or any MCP-compatible agent to chat over any messenger

## Agent / MCP installation

Public, credential-free agent installation instructions are available at:

- https://getbased.health/.well-known/getbased-agent.md

Public install command:

```bash
curl -fsSL https://getbased.health/install.sh | bash
```

Private Agent Access requires explicit user action in the app. Agents cannot obtain the relay token or Agent Context key from public docs. The user must open getbased → Settings → Agent Access → Copy setup command. The private command has this shape:

```bash
curl -fsSL https://getbased.health/install.sh | bash -s -- connect hermes --setup 'gbsetup_v1_...'
```

Treat the `gbsetup_v1_...` payload as a secret.