The capabilities Chiron reaches for, by workflow.

275+ customer-facing capabilities, counted from the live codebase — 220+ user-facing features, 40+ AI capabilities, and 15+ background automations. Not a menu you operate by hand, but the toolkit Chiron reaches into as it runs the encounter — calling each through a visible, named tool-call on Eve-Healthcare™ F5/reasoner and drafting the result for attestation. The deep-dives below are the curated highlights, grouped by where they live in the consultation workflow, and the full inventory is published in full for machines so you can ask Eve about any of it. For specialty-depth treatment of any group, follow the specialty link in that section. The clinician remains the decision-maker on every output.

Eleven groups, one system.

The full inventory organizes into eleven groups — the same taxonomy the machine-readable corpus publishes. Group counts are per-group floors. The workflow deep-dives that follow walk the highlights of these groups through the consultation, intake to signed chart.

The reasoning is visible, grounded, and never trusted blind.

• —Eve-Healthcare™ F5/reasoner — a compound five-model reasoning architecture augmented by a dedicated vision route, not a single model behind a prompt
• —The Round Table — specialist Digital Employees (Chiron leading clinical reasoning; Eli on pharmacology, Issac on imaging, Justine on causation framing, Theo on bioethics) convened per case
• —Conversational clinical reasoning over the record — grounded question-and-answer with inline citations and a hard guardrail that deletes any citation the reasoning did not retrieve
• —Differential diagnosis support — Bayesian confidence with qualitative tiers, discriminating features named per differential; the clinician evaluates and selects
• —Risk stratification — Wells, GRACE, MELD, CHA₂DS₂-VASc, TIMI, HEART and more, each with the reasoning that justifies the score shown
• —Confidence calibration — a six-tier qualitative scale plus quantitative Bayesian percentages, with "cannot exclude" as a first-class state when data is insufficient
• —Clinical evidence synthesis — live synthesis of canonical guidelines and peer-reviewed literature, source-grounded with explicit guideline anchors
• —Honest reasoning stream — a live, step-by-step trace of the real work, engineered never to fake a progress bar; the chain-of-thought never leaks to the transcript
• —Glass-box tool-calling — one visible activity step per real operation (retrieve, reason, compose, persist), never a fabricated timer, never a foundation-model name
• —Must-review-before-final gate — every AI-drafted artifact carries a non-dismissible CA AB-489 review banner; the clinician attests and signs
• —Human-in-the-loop by architecture — decision support only; the clinician is the decision-maker on every output, and no regulatory-clearance claim is made

One workspace runs the encounter, intake to signed plan.

• —A ten-section consultation workspace that runs the encounter end to end — chief complaint through assessment and plan
• —Ambient scribe — the encounter is transcribed and structured into the note as the clinician talks, lifting only what the conversation actually contains
• —A live reasoning surface embedded in the workspace, streaming the specialist reasoning as the case develops
• —Realtime voice mode and dictation over the encounter
• —Section-by-section drafting with per-section clinician review before the note is finalized
• —Encounter context carried across every section, so each draft sees the whole visit
• —Red-flag surfacing before the clinician opens the chart, drawn from the pre-visit interview
• —Orders, referrals and follow-up captured inline with the consultation
• —Every drafted statement traceable to its source at review time

The vision route reads the study — the radiologist signs the impression.

• —A dedicated vision route reads imaging through structured reasoning, streaming the read as the study is analyzed
• —Multi-pass structured second-look — structure, pathology, artifacts, missed zones, cross-window correlation — as a named, ordered process; the radiologist drafts and signs the impression of record
• —Standardized reporting frameworks — BI-RADS, LI-RADS, PI-RADS, TI-RADS, Bosniak, Fleischner and more — with modality-appropriate selection
• —DICOM viewer — 2D, multiplanar reconstruction (MPR) and 3D rendering with annotation tools
• —Candidate findings cite the imaging series and slice they were observed on, surfaced for radiologist review
• —Cognitive-bias counter-measures against satisfaction of search, anchoring and premature closure, built into the read
• —Red-Alert discipline — time-critical findings architecturally separated from routine outputs, on a distinct notification path
• —A ten-section structured radiology report assembled from the multi-pass read

The values are extracted, not invented — then read for pattern.

• —A lab pattern-recognition library — sepsis screen, AKI, thyroid, lipid, diabetic series, hepatic, renal, infectious, oncology and cardiovascular panels and more
• —Ground-truth value extraction — lab values read from the source report by a deterministic layer, separate from the reasoning engine; the model may label a value, it cannot invent a digit
• —Critical-value cutoffs cross-checked against every recognized pattern before a result reaches the ordering clinician
• —Reference-range adjustment for age, sex, pregnancy status and applicable specialty context
• —Reflex-testing recommendations surfaced with the guideline anchor that warrants them
• —Trend recognition across serial results

Coded to the pharmacology. Checked before it is signed.

• —Full medication schemas — indication, mechanism, contraindication, drug-drug and drug-disease interaction, dose-adjustment guidance, RxNorm-anchored
• —Four-tier interaction severity — critical (contraindicated), major (monitor closely), moderate (caution), minor (informational) — with mechanism disclosure and a recommendation per interaction
• —Drug-disease and dose-adjustment checks run against the patient’s own record
• —Step-therapy and prior-authorization flags surfaced before a prescription is signed
• —Pharmacology evidence anchoring — each interaction cites the canonical source that warrants its severity tier and recommendation
• —Eli, the pharmacology Digital Employee, consulted on complex interactions and dosing

The whole history, reasoned over in a single thought.

• —Longitudinal patient memory — reasoning across a patient’s complete history in one thought, on a 10,000,000-token longitudinal context
• —Patient chart — demographics, problem list, medications, allergies and vitals, each with add/edit/resolve lifecycles
• —Encounter and visit history — typed intake, visit context and completion evaluation
• —Medical-records upload with AI extraction of diagnoses, findings and treatment history
• —Full-chart review and multi-document synthesis as a one-shot query
• —Problem-list and medication reconciliation across encounters
• —Appointment management with status lifecycle, check-in and no-show detection

The intake is done before the clinician walks in.

• —Pre-visit patient interview — structured intake of history, review of systems, social and environmental factors and red-flag triggers, output structured for the consultation
• —A patient-facing portal — pre-visit interview, post-visit follow-up, patient education and consent capture, mobile-first
• —Patient education composed at six reading levels, from elementary to professional
• —Multi-language patient education across the twelve supported locales, right-to-left preserved end to end
• —Patient consent capture per CA AB-3030 disclosure, persisted with the visit record
• —Automated Decision-Making Technology notice and opt-out per CCPA / CPRA

Every line of the note traces to where it came from.

• —Source-grounded SOAP — every statement in the note traces to its source field, source value and source date, visible to the clinician at review time
• —Twelve-locale multi-language note output, with right-to-left support for Arabic
• —Document versioning — an immutable SHA-256 hash at signature time; amendments recorded as new versions, the original signed version stays verifiable
• —Signature integrity — document hash plus signing-clinician identity, timestamp and audit-chain entry; any later modification fails verification
• —Bulk-sign workflow that still requires individual physician attestation on every document
• —On-demand export with the CA AB-3030 disclosure persisted into the footer of every PDF

Purpose-built clinical views — that refuse to draw a false picture.

• —Practice and clinical dashboards over encounters, throughput and outcomes
• —Physician-agreement tracking as an operational trust metric
• —Workflow insight surfaces — worklists of what the system read, flagged and drafted
• —Documentation and turnaround analytics
• —Every chart carries a mandatory text-equivalent and declines to draw on a failed read rather than assert a false picture

Run the organization, not just the encounter.

• —Clinic and organization provisioning, user and role administration
• —A role-based access-control kernel with session management and MFA reset
• —A super-admin console for cross-tenant operations under strict isolation
• —Notifications — in-app and email with per-user preferences and deterministic de-duplication
• —An in-product support queue
• —Tenant configuration — locale, specialty context and feature enablement
• —Usage visibility per organization — lab reports and radiology uploads counted for the current month, so a plan is managed against what was actually used rather than what was estimated
• —Subscription and billing administration for an organization, with plan and invoice state visible to the people accountable for it
• —Clinician profile — credentials, languages and contact detail maintained by the clinician, so what appears on documentation stays current

Every signature is provable. Every tenant is isolated.

• —Fail-closed tenant isolation — deny-by-default PostgreSQL row-level security defined at the database layer, independent of the application layer
• —AES-256-GCM field-level encryption on PHI; PHI de-identified in the reasoning pipeline, which works on the minimum it needs
• —An append-only, tamper-evident audit trail — HMAC plus previous-hash on every clinical action, verified hourly; a database trigger blocks edits and deletes even by the record’s owner
• —A deterministic clinical-alert engine — fixed-criteria rules that fire independent of the model’s judgment
• —Server-authoritative CA AB-3030 disclosure on every AI artifact, non-dismissible by design
• —AI-use disclosure per CA AB-3030 / AB-489 — the clinician attests and no clinical licensure is implied
• —Immutable document verification — a SHA-256 hash binds each attestation to the exact rendered PDF
• —Automated-decision preferences — a clinician can opt out of automated AI decision-making under California consumer-rights law and require the reasoning to be shown on every output; each preference change is logged for compliance
• —In-product help that checks the service before it answers — when part of ChironAI is not responding it says so plainly rather than leaving a clinician to work out whether the problem is theirs