Huddle is Stream's pre-visit review surface. Open a patient five minutes before their visit and see what you've addressed in the last three visits, what's falling off your attention past ten months, and which HCCs you haven't captured this year — each with your own Assessment & Plan, confirmed, ready to read.
Most AI products applied to clinical charts just summarize everything. It's the obvious move — and the wrong one. What a clinician actually needs before a visit isn't a compressed version of the last eighteen months. It's a clear view of what's active, what's falling off, and what's outstanding. Huddle is what that view looks like.
Read Jake's full argument →Problems you've addressed in the last three visits. For each, Huddle shows the most recent Assessment & Plan — your words, as you wrote and signed them. If a problem appeared in all three recent visits, the first is visible by default; the others are a click away. You see the trajectory without re-reading three full notes.
Problems you haven't addressed in ten months or more. Huddle surfaces them so they don't stay forgotten. Each one comes with its last A&P — so you know exactly what state you left it in, and you can decide whether today's visit is the moment to pick it back up.
HCC-relevant problems you haven't captured this calendar year. Huddle flags them so nothing clinically legitimate slips past documentation. Capture happens as a byproduct of doing the visit — not as a separate admin chore.
Everything you see in Huddle is text you wrote and signed — or text a colleague on Stream wrote and signed, if you share patients. Nothing has been re-summarized. Nothing has been re-rendered through an LLM after the fact. When the chart tells you "last time, you adjusted her metformin because her A1c was 8.4," it's telling you exactly what you said, in the way you said it. This matters. It's the difference between reading a record and reading an interpretation of a record.
Tasks in Huddle aren't a separate list. They're attached to the medical problems they belong to, showing up under the relevant problem in every view. Complete them, snooze them, or bring them into today's visit — all from the Huddle screen. Stream auto-generates task recommendations from each visit and dedupes against existing tasks, so nothing gets added twice and nothing gets forgotten.
Huddle is possible because Stream's scribe produces a structured note from every visit — Subjective, Objective, and Assessment & Plan split out per medical problem. That structure is what lets Huddle surface the right A&P for the right problem at the right time, without re-running an LLM summarization layer over the whole chart. It's one architectural choice, and every downstream feature Stream builds depends on it.
See how the structured note works →Stream focuses on the entire patient — which is exactly what DPC needs. Progress over time matters more than any single encounter, and Stream is the only tool I've found that actually works that way.
$149/month for your first year. after. 30-day free trial. No credit card. Takes under three minutes to start.