Everything Northstar does.

The chart is FHIR-native end to end. Compositions, Conditions, AllergyIntolerances, MedicationRequests, Observations — every field is FHIR R4, every write is auditable, every read knows whose chart it's on.

The visit itself. Tiptap-grounded editor with AI ghost-text, AssemblyAI ambient scribe, peds-tuned SOAP drafting, and a sibling-shift detector that catches "let's talk about Mia now."

Keyboard-first quick entry. Live preview of derived metrics as you type. Batched as a single FHIR transaction.

Standard percentile curves done right, plus the interpretation layer that turns the chart from a record into clinical guidance. CDC, WHO, Fenton 2013, Olsen 2010 — each cited.

Pediatric immunization is half the job. The vaccine engine is encoded data — schedule, intervals, catch-up — with a Cmd-K-fast search, lot tracking, and atomic FHIR transactions.

Bili and dosing — the two safety-critical calculators every peds practice runs daily. Both source-cited, both clinician-triggered, both audit-captured.

Developmental and behavioral screening done as a first-class longitudinal tracker — not a checklist that resets per visit.

Chronic conditions tracked across visits, not relived per encounter. FHIR CarePlan with starter templates and source-grounded AI delta suggestions.

After-visit summaries, forms, action plans, anticipatory guidance — all generated source-grounded, rendered on letterhead, persisted as DocumentReference, sent via Spruce or downloaded.

First-class siblings, custody-share guarantors, half-sibling taxonomy. The data model knows pediatrics is a household business.

One workspace for Spruce, Gmail, faxes, staged documents, forms, and labs. AI-triaged on every ingest path. Same-page focus selection. Refill resolution in one click.

The everyday surfaces: today's home screen, the patients list, the schedule grid, and the operational rails underneath.

Northstar plays well with the tools your practice already runs. Patient messaging, email, drive, calendar, ambient transcription — all wired in.

Seventeen AI surfaces, each doing something only AI can do — and each refusing to do what's unsafe. This is the complete map of what the model actually writes on your behalf.

Every AI surface ships with a strict prompt + a parser that refuses unsafe output + a judge that scores it + graded reference scenarios. Nothing gets shipped that we can't grade.

HIPAA-grade audit. Auth boundary on every PHI route. Token envelope encryption. Sanitized HTML on note writes. Production-readiness gates.

A modern, opinionated layer. Built to drop in, evolve fast, and never trap your data.