Your health data,
organized for you
and your AI.

MCP (Model Context Protocol) is an open standard that lets AI assistants like Claude connect to external data sources. twyn gives you a personal MCP URL — paste it into Claude Desktop or claude.ai once, and your agent can read your health data: structured, current, and scoped to you alone.

Yes. Ask twyn is a built-in coach: open it from any screen and ask in plain language — "what should I focus on?", "what changed since last quarter?", "is this supplement right for me?". twyn reads your stack through the same tools an external agent would, knows which screen you are on, and writes back (saving notes, updating labs, setting recommendation stances) only after you approve. It runs on Gemini by default for cost and latency, with Anthropic available for premium reasoning.

No. Ask twyn is included on the Coach plan — twyn provides the model and absorbs the inference cost. If you prefer to run your own agent (Claude Desktop, Cursor, claude.ai web, Claude Code), the Connect plan gives you a personal MCP URL with the same structured read and write access. Most users start with Ask twyn and add MCP later if they want their agent to handle health alongside everything else.

Ask twyn is built in: one click from any screen, screen-aware (it knows you are looking at ApoB, or this week's sleep), streaming, with click-to-approve actions. Connecting your own agent via MCP is for power users who already live in Claude Desktop, Cursor, or claude.ai — your agent gets the same structured access from outside the app, but loses screen context and the inline approval cards. The data surface is identical; the difference is friction and where the conversation happens.

A chatbot can read a pile of files once, but it does not become a reliable health memory. twyn turns labs, wearables, vaccines, supplements, DNA, notes, and records into a structured source of truth — with dates, units, provenance, freshness, and history — so your AI asks the data instead of re-parsing a messy archive every time.

In a managed Postgres database (Supabase) with per-user row-level isolation, encrypted at rest and in transit. Raw files you upload — lab PDFs, exports — are preserved unchanged in private object storage, so your data is always re-processable and never locked into our parsing.

Only you, and any AI agent you explicitly connect with your personal access token. Tokens are issued and revoked by you in Settings. When you use AI features (Ask twyn, lab extraction, note structuring), relevant excerpts are processed by Anthropic or Google under their API terms, which exclude training on your data.

twyn requests only the OAuth scopes needed to sync the wearable data you choose to connect. For Oura, that can include sleep, readiness/activity, heart-rate, workouts, sessions, tags, SpO2, and profile data. For WHOOP, that can include recovery, cycles/strain, workouts, sleep, profile, and body measurements. Provider tokens are stored encrypted, used only for sync, and can be disconnected from Settings.

Yes — your data is yours. In the app, Settings → Privacy & Data lets you download a complete export or permanently delete your account and health data. For anything else, email hello@twyn.life.

No. Ask twyn and AI-powered extraction are included. If you prefer running on your own Anthropic key (heavy use, custom model, enterprise terms), you can plug it in from Settings — useful for power users on the Connect plan running their own agent.

Anything that speaks MCP: Claude Desktop, claude.ai web connectors, Claude Code, Cursor, and any other MCP-compatible client. You paste one URL with your personal token and your agent sees your whole stack. The in-app Ask twyn coach does not need any external client — it works the moment you sign in.

No. twyn is a system of record: it aggregates, normalizes, and exposes your data. It does not diagnose, prescribe, or replace a clinician. What it does make dramatically better is the quality of conversations you have — with your AI and with your doctor.