Connect your AI
Hemix · hemix.life · Use your own AI assistant with your Hemix data
Hemix is also a Model Context Protocol (MCP) server. That means you can connect your own AI assistant — Claude, ChatGPT, Cursor, or any MCP-capable tool — directly to your Hemix record. With your permission, it can read your results and answer questions in plain language. Reads are read-only; and, if you allow it, your AI can also propose additions — a supplement, a journal note, a reminder — that you approve before anything is saved. It can never write to your record on its own, and you can disconnect at any time.
Connecting your own AI is rolling out gradually. If it isn't enabled on your account yet, you'll see a notice when you try to connect.
What you'll need
- A Hemix account with some results in it.
- An AI tool that supports custom MCP connectors (for example, Claude).
- About two minutes.
Connect in three steps
In your AI tool, add Hemix as a custom connector using this address:
- In your AI tool, open Settings → Connectors → Add custom connector.
- Paste the address above and add it.
- Hemix opens a consent screen — choose what to allow: read access, and optionally let it propose additions you'll confirm (supplements, journal notes, reminders). You'll return to your tool, connected.
You only sign in and approve once. After that, your tool can read your record whenever you ask it to. Step-by-step guides: Claude, ChatGPT, Cursor & VS Code.
What your AI can read
Once connected, your AI can look up the following from your record. Each is read-only:
- Your profile — the demographic details used for age- and sex-aware ranges.
- Your biomarkers — your most recent results, and the full history of any single marker over time.
- What's outside range — the markers in your latest results that fall outside the reference range.
- Your supplements — what you log, with doses and dates.
- Your journal — your logged lifestyle entries.
- Your wearable data — which metrics are available, the day-by-day history of one (steps, sleep, …), and your workouts.
- A date in context — everything recorded around a chosen date — labs, supplements, wearables — side by side.
- The public catalog — reference definitions and ranges for any marker (this is reference data, not your results).
On a free account, your AI sees a recent slice of your history; a Pro account exposes the full series and logs.
What your AI can propose — you confirm everything
If you grant write permission, your AI can also propose additions to the parts of your record you author yourself:
- A supplement or medication you're taking — and updating its dose, or marking it ended.
- A journal / lifestyle note you dictated.
- A reminder — e.g. "retest ferritin in three months" — or rescheduling one.
Every one of these is a proposal, never a direct write. Your AI hands you a Hemix confirmation link; you open it, see exactly what will be saved, and approve — only then is it stored, tagged as added by your agent and editable anytime. Write permission is granted separately from read (per category: supplements, journal, reminders) and can be revoked independently. Hemix records what your agent stored; it never authors recommendations.
What it can't do
- It can't write to your record on its own. Writes are proposals only — nothing is saved until you confirm it on Hemix. Reads never change anything.
- It reads your data, not Hemix's opinion. Hemix hands over your stored values, units, ranges, and dates. Any interpretation is produced by the tool you connect — Hemix doesn't send a verdict.
- It only sees your record. The connection is scoped to your account and to any family members who have consented — nothing else.
You stay in control
- You approve the connection yourself, and every connected tool is listed in your Hemix settings.
- You can disconnect any tool instantly — access stops immediately.
- Hemix records when and which record type a tool read — counts and times only, never your values.
- Your data stays EU-hosted, and each connection is scoped to your own record.
Try asking
Once connected, you can ask your AI things like:
- "When did my ferritin start trending down — and what else changed at the same time?"
- "Compare my lipid panel from 2024 to today. Where am I better, where am I worse?"
- "Build me a quarterly brief I can take to my next physician visit."
Technical reference
For developers and technical users:
transport Model Context Protocol over HTTP
auth OAuth 2.1 + PKCE · consent at hemix.life/connect
read access RLS-scoped read-only · scoped to your account
write access propose→confirm only · scopes write:supplements / write:journal / write:reminders · committed on hemix.life/confirm, tagged written_by=agent
read tools get_profile · get_context_manifest · list_biomarkers · get_biomarker_history · list_out_of_range · search_catalog · get_supplements · get_lifestyle_log · get_context_for_date · get_wearable_summary · get_wearable_history · get_workouts · list_reminders
write tools add_supplement · update_supplement · end_supplement · add_journal_note · add_reminder · reschedule_reminder · complete_reminder
Read tools return your stored data with a provenance marker; none returns a Hemix-computed score or interpretation. Write tools create a pending draft and return a confirmation link — the agent never commits; you do.
A note on what you get back
Hemix is a wellness tracking tool, not a medical device. Educational content is not medical advice.
Hemix provides your data; it does not provide a diagnosis or medical advice. Anything your connected AI tells you is produced by that third-party tool, which you control. Discuss any changes in your health with your physician.