Four apps, and none of them know the other three exist.
You don't eat badly because you don't know what's good for you. You eat badly because the plan lives in four places. Your workout app has no idea you slept four hours. Your sleep tracker has no idea you mentioned a headache to a doctor three months ago that is happening again right now. Nobody coordinated any of it, because coordinating it was never anyone's actual job — including yours, even though it's your body.
WITHOUT A TEAM
Four separate records and no shared memory. A tracker holds whatever you type into it and does nothing with it — it won't notice Tuesday's fatigue sits right next to Sunday's sleep debt. You have to notice that yourself, which is the job that was already failing.
WITH HEALTH TEAM
Every agent writes into the same record, so the pattern is already assembled by the time anyone reads it. Close the tab and nothing is forgotten — the memory outlives the session, which is exactly what a regular chatbot can't do.
This one actually runs. Log something and watch it move.
Everything below is live state. What you type routes to an agent, changes that agent's card, lands in Today, and shows up in Sunday's digest — the same handoffs the real build makes.
0 ENTRIES THIS SESSION
Type anything — a meal, how you slept, a symptom, a workout. The Chief routes it to the right agent automatically.
RULE-BASEDDemo only — nothing you type is stored, sent anywhere, or evaluated. Routing is keyword-based so you can watch the Chief hand work off.
5h 40m last night. Third night under six hours in seven days.
Breakfast photographed and tagged in four seconds. Protein 18g — third morning under your usual.
Heavy lower body swapped for 30 minutes of mobility. No streak broken, no notification about missed goals.
Refill due in three days. Pharmacy pickup window added to your week.
Sample day, plus anything you log above. Everything here except your own notes happened without a prompt.
One pattern worth your attention. Everything else is noise this week.
- 01All three headaches this week landed the morning after a night under six hours. That connection sat across three different agents and would have taken weeks to spot by hand.
- 02Protein came in under target on three of five logged days, tracking the same short-sleep nights. Flagged, not yet linked to anything.
- 03Two training sessions were missed and the week was rebuilt around them. Weekly volume held.
- 04One supplement interaction is waiting on your pharmacist to confirm before the refill goes through.
WORTH MENTIONING AT YOUR NEXT APPOINTMENT — NOT WORTH A CALL TODAY.
Turn AGENT AUTONOMY off and the agents stop rebuilding your week on their own — try it, then log a short night. One switch here does not move at all.
NAME THE SIX
Setup is job titles, not code. You brief an agent the way you'd brief someone on their first day.
CONNECT SOURCES
Watch, photos, voice notes, prescriptions. Each agent reads only what its job needs.
LOG CARELESSLY
Type it however it comes out. The Chief routes it. You never pick an app again.
READ SUNDAY
One digest. If something needed you sooner, you already heard about it.
Six specialists, one chief, and a memory none of them lose.
The failure mode of a single health bot is that it pretends to be a trainer, a dietitian and a scheduler all at once, badly. Here each agent has one job, one lane and one voice — and a chief that reads all of them so you don't have to.
THEY LOG
Meals, sleep, sessions, doses and symptoms land in one record instead of four apps.
THEY ACT
Inside its lane an agent doesn't ask permission. Workout Plan just rebuilds the week.
THEY CROSS-REFERENCE
Short sleep on Tuesday and a headache on Thursday stop being two unrelated notes.
THE CHIEF REPORTS
One Sunday digest of what actually deserves your attention, and nothing else.
Symptom Triage never reassures and never guesses. That one is not supposed to be clever.
Every agent handles what it can on its own — Meal Log doesn't need permission to point out low fiber. The one hard boundary is Symptom Triage's ceiling: mild things get logged and trended quietly, but anything on the red-flag list reaches you the second it's reported. No judgement call, no "probably fine."
That rule exists because people trust an AI's bedside manner over an actual doctor's. Health Team does not replace a doctor. Symptom Triage's entire job is making sure you still go see one — and it's the only setting in the product you can't turn off.
Nothing here needed a prompt except one photo.
- Third short night flagged
Five hours forty. The watch synced before you were up, and the number went into the same record everyone else reads.
- Breakfast tagged in four seconds
One photo. Protein came in low for the third morning, so it was flagged and parked rather than announced.
- Heavy session swapped for mobility
It read the sleep number, not the calendar. No passive-aggressive notification about a missed goal.
- "dull headache again"
Typed into what was meant to be a note for Meal Log. The Chief routed it to the right agent anyway.
- Logged, linked, not evaluated
Checked against the red-flag list, found nothing on it, and linked the note to two other headaches from the same week.
- The pattern lands in one line
All three headaches arrived the morning after a short night. That connection lived across three agents and would have taken weeks to see by hand — if you ever saw it at all.
Three people who'd normally split this job and never talk to each other.
A PERSONAL TRAINER
Who plans your week on Monday and never learns you slept four hours on Wednesday. Workout Plan reads the sleep before it reads the schedule.
A NUTRITIONIST
Who sees you monthly and works from what you remember. Meal Log works from a photo taken in four seconds, every day, and watches the trend.
A CONCIERGE FRONT DESK
Who handles refills and appointments and nothing else. Med & Appts does that, and also checks what you added next to an active prescription.
None of those three ever coordinated with the other two — that was never anyone's job. What Health Team actually replaces is the unpaid, exhausting work of catching the pattern yourself in time to bring it up.
What the team will not do, on purpose.
- It does not diagnose. Naming what's wrong with you is a clinician's job.
- It does not change a prescription, a dose or a treatment plan.
- It flags a possible interaction for your pharmacist. It never rules one out.
- It does not soften, contextualise or trend a red-flag symptom.
- It is not a medical device and nothing in it is medical advice.
- In an emergency it names the next step and gets out of the way.
The ones worth asking first.
Why six agents instead of one assistant?
What does the Chief actually do?
Do I have to log things in the right place?
What happens if I report something serious?
Is this a replacement for my doctor?
Where does my data live?
BRIEF YOUR SIX
Name the agents, point them at your own data, and read one digest on Sunday instead of six apps all week.