medOS ultra

The generational leap in hospital software

Your hospital has outgrown its software.

First-generation clinical platforms were built for a different decade — slow to change, expensive to run, and closed by design. medOS ultra is the modern, AI-native hospital operating system that deploys in weeks, adapts to any market, and speaks open standards from day one.

Weeks to first live departmentAI-native, not bolted onFHIR R4 + HL7v2 in the core17 locales · any market

The cost of standing still

What a legacy platform quietly charges you

The licence is only the beginning. The real price is paid in time, flexibility and missed intelligence — every day the old system stays in place.

Implementations measured in years

Multi-year rollouts, high-stakes big-bang go-lives, and seven-figure consulting bills before a single patient is seen.

Every change is a change request

Workflows, clinical rules and forms are locked inside vendor code. A new regulation or a new clinic means a ticket, a quote, and a wait.

Intelligence sold as an add-on

AI arrives years later as a premium module — isolated from the chart and the workflow, priced per seat, and hard to trust.

Built for a single market

Local insurance schemes, languages and regulatory packs are costly rebuilds instead of configuration you control.

Side-by-side

Legacy platforms vs. medOS ultra

The same job, two generations apart. Here is what changes the day you switch.

Legacy platforms
medOS ultra
✓ Recommended
Time to go live
18–48 months, big-bang cut-over
Weeks to months, department by department
Total cost of ownership
Seven-figure licences + consulting
Open foundation — a fraction of the cost
Built-in intelligence
Extra-cost module, bolted on
AI cowork substrate, recommender-first
Changing a workflow or rule
Vendor change request, months
Self-service data rows, live in minutes
Interoperability (FHIR / HL7)
Paid add-on, partial coverage
Native FHIR R4 + HL7v2, 80+ connectors
Cross-department updates
Nightly batch / manual refresh
Realtime event mesh on every screen
Multi-country & languages
Costly per-country rebuilds
Market packs · 17 locales · any region
Technology foundation
Proprietary, decades-old stack
Modern open stack, 16 microservices
Deployment
Vendor cloud or heavy on-prem only
Cloud, on-prem or air-gapped — your call
Modularity
Monolith — pay for all of it
Install only the modules you need
Clinician experience
Dense screens, training-heavy
Guided, realtime, mobile-ready
Upgrades
Disruptive, costly projects
Continuous and additive — no big-bang
Your data
Locked in a proprietary format
Open standards — portable, and yours

Built to be extended

Plug in what you need. Nothing you don’t.

medOS is assembled, not bought whole. Every capability is a module that snaps onto a shared foundation — and the things you'd file a change request for elsewhere are data rows you edit yourself.

Plug & playAI cowork substrateonLab & ImagingonPharmacy & e-MARonClinical SuiteonFoundation · master datacoreOncologysnap to install

Install only the modules you need

Clinical, pharmacy, lab, blood bank, oncology, nuclear medicine — each is a self-contained module. Toggle one on and the installer resolves its dependencies, seeds its data and lights up routes. Pay for what you run.

A new country is a market pack, not a fork

Local insurance schemes, rates, terminology and locale ship as a market pack you drop in. One codebase serves the US, UK/Europe, Japan, China and SE Asia — 17 locales, zero forks.

Workflows, rules & billing are data rows

CDS rules, policy gates, workflow templates and facility billing live as editable rows — change them live in an admin screen. A new regulation or clinic is a configuration, not a vendor release.

module.jsonVITE_ENABLED_MODULESdependency-resolving installer
module installer — plug in, configure as data, no redeploy
MODULES · install only what you needVITE_ENABLED_MODULES

Clinical Suite

requires: foundation

installed

Pharmacy & e-MAR

requires: clinical

installed

Lab & Imaging

requires: diagnostic

installed

Blood Bank

requires: clinical

installed

Oncology & Chemo

requires: clinical · pharmacy

+ install

Nuclear Medicine

requires: imaging · RIS/PACS

+ install
CONFIGURE AS DATA · live, no code editable
cds_rulesrow · effective now

NEWS2 ≥ 7 → critical

policy_gatesrow · effective now

discharge — settle balance first

facility_billing_rulerow · effective now

🇯🇵 kaigo per-unit × region ¥

workflow_templatesrow · effective now

ER → triage → bed → orders

Toggle a module → the installer resolves dependencies, seeds its market-pack data and lights up routes. Adapt a clinic, scheme or country by editing rows.

Every graphic on this page is the real product surface, drawn in the browser — not a screenshot. That is the point: the system is built from composable pieces all the way down.

What you gain

Six reasons teams never look back

Not a feature checklist — a different operating model. Intelligence, flexibility and openness built into the foundation, not sold back to you later.

Intelligence

AI-native, not bolted on

Coder, nurse, pharmacist and RCM coworkers operate as named identities. Every proposal lands in a human inbox — accept, edit or reject. No autonomous clinical writes, ever.

Flexibility

Configure by data, not code

Workflows, CDS rules, policy gates and billing are live-editable data rows. Adapt to a new clinic, scheme or regulation without waiting on a vendor release.

Operations

Realtime by default

Every order, result, bed move and ledger entry is a hospital event on a NATS mesh, projected into realtime read-models. No nightly batch, no polling, no blind spots.

Global

Built for every market

Market packs carry local schemes, rates and locale. One codebase serves the US, UK/Europe, Japan, China and SE Asia — a new country is configuration, not a fork.

Interoperability

Open standards in the core

FHIR R4 read/write + subscriptions, HL7v2 ADT/ORM/ORU over MLLP, DICOM worklists, and a connector catalog of 80+ integrations. Connect labs, devices and partners on day one.

Financials

Revenue you don’t leave behind

An append-only billable ledger captures every stock issue and procedure at a resolved price. ER, IPD, OR and LR write to the same spine; country rule packs handle the claims.

Migration, simplified

Connect what you have. Go live the same day.

Migration sounds like a year-long project. It isn't. HDAP — the Data Activation Hub — connects to the systems you already run and streams them into medOS, live. You're not moving data; you're turning it on.

SYSTEMS YOU ALREADY RUNONE LIVE PLATFORMEHR / HISHL7v2 · FHIR R4Laboratory / LISHL7v2 ORUImagingDICOM · RIS / PACSERP / HRMOdoo · SAP ODataClaims & payerE-Claim · ECLIPSE+ Your next systemany of 80+ connectorsHDAPData Activation Hubconnect · normalize · stream20 categories · 80+ connectors · realtimemedOS ultralive · realtime read-modelsone chart, one ledger, one twin
1

Connect

Point a connector at a system you already run — HL7v2, FHIR, DICOM, an ERP feed, a CSV. No data migration, no export project. It stays where it is.

2

Activate

HDAP normalizes the feed and streams it onto the event mesh. One hub does the mapping once — not a brittle point-to-point integration per system.

3

Live

It shows up in medOS in realtime — same chart, same ledger, same twin. Most systems are live the same day. Add the next one whenever you're ready.

That’s the whole pattern. One hub, connect once per system, live in realtime — no rip-and-replace, no data-migration weekend, no integration spaghetti.

Migration, de-risked

A switch that never asks you to leap

No rip-and-replace. medOS runs alongside your current system, mirrors it in realtime, and lets you move one department at a time — fully reversible at every step.

migration — parallel run · shadow mode · reversible

Incumbent system

stays the system of record

FHIR · HL7
realtime mirror

medOS ultra

mirrors, then takes over

Cut over one department at a time
OPD · Registrationcut over
Pharmacycut over
Laboratoryvalidating in parallel
IPD Wardsshadow mode
Operating Roomsconnected, not yet live

Reversible at every step. No big-bang weekend, no war room — roll a department back to the incumbent any time until you’re sure.

Four steps, zero downtime

01

Connect

Our FHIR & HL7 adapters sit alongside your current system. Nothing is switched off, nothing is at risk.

02

Mirror

Live encounters, orders and results populate medOS read-models in realtime. Your teams watch the data flowing in.

03

Run in parallel

Operate medOS in shadow mode next to the incumbent. Validate every workflow and report with zero downtime.

04

Cut over

Move department by department, on your schedule. No big-bang weekend, no war room — just steady, reversible progress.

The business case

Run your own numbers. See the payback.

A legacy renewal is a seven-figure decision made on a vendor's spreadsheet. Here's one you can drive yourself — move the sliders to your estate and watch the five-year gap open up.

400
$1.2M

Assumptions (Typical): medOS run-rate ≈ 47% of today; go-live ≈ 35% of one year’s spend (parallel-run, weeks); legacy major upgrade in year 3. Illustrative model — we’ll run your real figures in a migration review.

$4.2M
5-year savings
53%
Lower annual run-rate
8 mo
Payback period
CUMULATIVE COST · 5 YEARSStaySwitch
Y1Y2Y3Y4Y5legacy re-platformpayback 8 mo$7.4M$3.2M

Drag the sliders. The shaded gap is what you keep by switching — this estate clears payback in 8 mo.

The other half of the case

The cost is the easy win. The revenue is the bigger one.

A legacy renewal only ever talks about what you spend. The harder number is what you never collect — the charges that leak, the codes left on the table, the claims quietly denied. medOS turns the same AI that runs the floor onto the bill, and gives that revenue back.

Revenue Explorer

Potential revenue with AI — four levers, four real surfaces. Move the sliders to your estate and watch the recoverable revenue add up.

$80.0M
45%

Assumptions (Typical): charge capture 1.5%, denial reduction 1.0% and reimbursement 0.6% of total revenue; coding 2.0% of inpatient revenue. 5-year figure applies a year-1 adoption ramp, then modest compounding as the AI calibrates on your data. Illustrative model — we’ll run your real figures in a revenue review.

$3.2M
Annual revenue uplift
+4.0%
Of gross revenue
$15.7M
5-year cumulative
WHERE THE MONEY COMES FROM$3.2M/yr recovered
Charge capture recoverybillable_ledger38%$1.2M
Coding accuracy (CC / MCC + specificity)AI medical coder22%$720k
Denial reductionRCM validation25%$800k
Reimbursement optimizationHORUS reimbursement agent15%$480k

Drag the sliders. Every lever is a real surface — recommender-first, a clinician or coder approves each proposal before it bills.

Charge with AI — the medical coder

medos — coding worklist · AN 66-04211 · cowork: coder

DRG 195 · Simple pneumonia

coded $11,240 · status: pending review

3 AI proposals
Missing MCC

J18.9 Pneumonia

+ J96.01 Acute respiratory failure

Documented in the progress note, never coded — lifts the DRG.

+$4,180
AcceptEditRejectlogged · llm_audit_log
Specificity

I50.9 Heart failure, unspecified

→ I50.43 Acute-on-chronic diastolic HF

Echo + meds support the specific code. Higher case-mix weight.

+$1,920
AcceptEditRejectlogged · llm_audit_log
Charge leakage

2 × infusion set issued from ward stock

consumed · never charged

billable_ledger flags the gap between stock issued and lines billed.

+$340
AcceptEditRejectlogged · llm_audit_log
Recoverable on this encounter+$6,440

The coder approves each line — no code is ever changed autonomously. Every accept, edit and reject is the training label.

The coder coworker reads the chart and proposes the missing complication, the more specific diagnosis and the supply that was used but never billed — each one a line a human accepts, edits or rejects. No code is ever changed on its own, and every decision is the label the model learns from.

Revenue at risk — powered by HORUS

horus — twin · reimbursement · revenue at risk
GROSS CHARGES → NET COLLECTEDAI recovers 5.1% of gross
1001.5recovered2.0recovered1.6recovered94.9Gross chargesCharge leaka…Under-codingDenialsNet collected

The dashed bars are revenue most hospitals quietly write off. medOS closes each gap before the claim leaves — the reimbursement agent flags revenue at risk before discharge, not at month-end.

The HORUS twin keeps a running picture of revenue at risk — leakage, under-coding and likely denials — and the reimbursement agent flags it before discharge, routing each charge to its best-paying eligible scheme. You close the gap while the patient is still in the bed.

Same governance, applied to money. The revenue AI is recommender-first, audited per inference, and off until you switch it on — it proposes, your coders and clinicians decide.

Your data, your call

The opposite of lock-in.

You're leaving a platform because your data was held hostage in a format only the vendor could read. medOS is built so that can never happen to you again — the exit door stays open by design.

data portability — open in, open out, leave any time

Your medOS data

charts · ledger · events · documents

Exports as

FHIR R4 BundleFHIR Bulk $exportHL7v2Full DB access (on-prem)

Goes anywhere you want

Self-host on-prem

your hardware, full DB access

Any FHIR system

another vendor, your call

Load your warehouse

from the $export NDJSON

The exit door is never locked. No proprietary format, no extraction fee, no hostage data.

Open standards, both directions

FHIR R4 and HL7v2 read and write in the core, DICOM via connector — your records are never trapped in a proprietary schema you can't read.

Export on your terms

Pull your estate out through FHIR R4 Bundles and the FHIR Bulk Data $export (NDJSON); on-prem deployments keep full database-level access. No extraction fee, no gatekeeping.

Self-host the foundation

Run the foundation on your own hardware and keep operating even if you never speak to us again. Source access comes through the apply-first foundation program.

A real exit strategy

Leaving is a supported path, not a threat. The thing you're escaping — lock-in — is the one thing medOS is architecturally designed not to do.

Security & sovereignty

Safe enough to bet the hospital on

Nobody signs off on a hospital platform for its features — they sign for the absence of a breach, a leak, or a lock-in. medOS is built to be inspected, deployed where you control it, and audited end to end.

deployment — your infrastructure, your jurisdiction, your call
Cloud

managed, per-region

AWS / GCP, your account or ours

in-region residency

auto-scale + DR

On-prem

your data centre

single Docker Compose stack

your hardware, your network

no data leaves the building

Air-gapped

fully isolated

runs fully offline

offline model serving

isolated network

Same codebase, same features in every mode. Data residency is a deployment choice, not a premium tier.

Behavioral threat detection (RUDS)

Platform-wide anomaly detection — credential stuffing, bulk export, off-hours spikes, snooping — scored inline (<50ms) and nightly, wired to real-time alerts.

Role & plane-scoped access

Read models carry row-level security, and every AI coworker is confined to a plane-scoped, consent-gated surface — the back office never leaks into patient-facing views.

Data residency you control

Cloud, on-prem or fully air-gapped. Your data stays in your region and your jurisdiction — residency is a deployment switch, not an upsell.

Audited end to end

Every action, every API call and every AI inference is attributable and logged, with BAA / DPA available. Tamper-evident anchoring is on the roadmap.

HIPAAGDPRPDPAAPPI-readyFHIR R4HL7v2Air-gap capableSelf-hostable

Deployments are designed for these regimes and standards. Certification scope is confirmed per deployment.

Responsible AI

AI that proposes. Humans that decide.

The first question anyone asks is how the AI is governed. The answer is simple and absolute: it never acts on its own. Every output is a proposal a named clinician accepts, edits or rejects — and every one is audited.

ai-governance — recommender-first, human-in-the-loop, audited

AI coworker proposes

Coder · nurse · pharmacist · RCM. Named identity, plane-scoped data grants.

Human inbox

AcceptEditReject

Clinician disposes. The gesture is the training label.

Audit log

Every proposal + decision recorded. Per-inference, attributable.

Invariant · No autonomous clinical writes — ever
Invariant · AI cannot raise severity or open a gate
Invariant · Per-tenant kill switch + off by default

Proof, not promises

One codebase, already live across markets

This isn't a single-hospital pilot stretched into a brochure. The same platform runs live across regulatory regimes and languages today, with more markets packaged and ready — a new country is a market pack, not a rebuild.

🇹🇭

Thailand

NHSO / SSO · 16-file E-Claim · th locale

Live
🇯🇵

Japan

Kaigo LTC · hospital + nursing home · ja locale

Live
🇵🇭

Philippines

PhilHealth case rates · fil locale

Live
🇨🇳

China

market pack + zh locale

Ready
🇺🇸

US

FFS / DRG · X12 claims-ready

Ready
🇬🇧

UK / Europe

GDPR · FHIR-first

Ready

17 locales, multi-region by configuration. Local insurance schemes, rates and regulatory packs are data you control — the clinical engine underneath never forks.

The upside

Not a cost to contain — an advantage to compound

A legacy renewal buys you another five years of standing still. medOS turns the two assets you already own — your operations and your outcomes — into an intelligence advantage that gets stronger every day.

HORUS twin

Decisions, not just dashboards

A digital twin of the hospital — capacity, demand, workforce, supply, reimbursement — runs 7 agents over DES + Monte-Carlo to pressure-test a decision before you make it, calibrated from your live read model.

Outcome corpus

Built to learn from your own outcomes

A governed pathway turns your encounters, orders and results into a de-identified corpus your models can calibrate on — your population, not a generic benchmark from someone else's hospital.

Compounding

The longer you run it, the sharper it gets

Recommendations, forecasting and coding are designed to improve on your data. That advantage compounds, it stays in your region, and — like everything here — it's yours.

The hard questions

Everything a skeptic will ask

The objections you should raise — answered straight, no hand-waving.

What if you go out of business?

medOS is self-hostable on your own infrastructure — you can run, operate and extend it independently, with source access through the apply-first foundation program. Your operations never depend on us being around.

How do we integrate the systems we already run?

Through HDAP and a connector catalog of 80+ integrations across 20 categories — HL7v2, FHIR R4, DICOM, ERP/HRM feeds, lab instruments, claims rails. You connect what you have rather than replacing it.

Who's accountable for the AI?

A clinician is. Every AI output is a proposal that a named human accepts, edits or rejects — there are no autonomous clinical writes. Each inference is audited and attributable, and the whole AI layer is off by default with a per-tenant kill switch.

What about downtime and disaster recovery?

Event-driven services on a NATS mesh with realtime read-model projections, health checks and ordered startup. Deploy cloud (auto-scale + DR), on-prem, or air-gapped — your reliability posture, your choice.

How long does it really take?

Department by department, in weeks — not a multi-year big-bang. medOS runs in parallel with your incumbent, mirrors it in realtime, and you cut over on your schedule, reversibly.

Is our data locked in?

No. Open standards in and out, export on demand via FHIR R4 Bundles and Bulk $export, plus full database-level access on on-prem deployments — no extraction fees. Lock-in is the one thing the architecture is explicitly designed to prevent.

Weeks

Not years to first live department

16

Microservices on one event mesh

17

Locales, multi-region by data

80+

Connector integrations in the catalog

100%

Open standards — your data stays yours

Zero-risk first step

Prove it on one department. Risk nothing.

We’ll stand up a parallel-run pilot for a single department — your real workflows, mirrored live next to your incumbent. If it doesn’t prove out, roll it back. You’ve lost nothing but a few weeks of seeing the future.

No commitment · no big-bang · fully reversible · your data stays yours

Ask Anything