One fleet of AI agents does the interface work your team can't hire fast enough for
Health1st is not a single model bolted onto an integration engine. It is 12 specialized agents — each owning a domain, each with human-in-the-loop review and full audit — that map, translate, validate and reliably deliver healthcare data across HL7 v2, FHIR, X12 and NCPDP.
- 12 domain agents, one orchestration layer
- Bidirectional by design: v2 ↔ FHIR ↔ X12 ↔ NCPDP
- Every action versioned, explained and audited
MSH|^~\&|EPIC|HOSP|H1ST|DST|202601..||ADT^A01|
PID|1||MRN12345^^^HOSP^MR||DOE^JOHN^A||
19701225|M|||123 MAIN ST^^AUSTIN^TX
PV1|1|I|ICU^101^A|||...admit...
DG1|1|I10|E11.9^Type 2 diabetes mellitusAgent
{ "resourceType": "Patient",
"identifier": [{ "value": "MRN12345" }],
"name": [{ "family": "Doe",
"given": ["John","A"] }],
"gender": "male",
"birthDate": "1970-12-25" }
{ "resourceType": "Condition",
"code": { "coding": [{
"system": "…/icd-10-cm",
"code": "E11.9" }] } }In short
/platform/ai-agents#mapping.A single message, handled end to end
No agent works alone. The orchestration layer hands each message down the fleet — from onboarding a new partner to delivering a validated, conformant result — and heals failures automatically.
- 1
1 · Onboard & learn
The Onboarding Agent ingests sample messages from a new EHR, lab, pharmacy or payer, discovers endpoints and bootstraps the first interfaces — days, not weeks.
- 2
2 · Map & normalize
The Mapping Agent drafts field/segment mappings (incl. Z-segments) while the Terminology Agent crosswalks ICD-10, SNOMED, LOINC, RxNorm and NDC and binds value sets to US Core.
- 3
3 · Transform & validate
The Transformation Agent executes the bidirectional conversion with round-trip fidelity; the Validation Agent confirms conformance to US Core / Da Vinci / IPS before anything ships.
- 4
4 · Route, secure & watch
The Orchestration Agent guarantees delivery with retry/replay, the Compliance Agent enforces consent, provenance and audit, and the Observability Agent watches every message for anomalies.
The same choreography runs in reverse (FHIR → HL7 v2) and at any scale — real-time, batch, streaming, or Bulk FHIR $export for population jobs.
The translation core (live MVP wedge)
The agents that do the heavy lifting of bidirectional HL7 v2 ↔ FHIR translation — mapping, terminology, conversion, validation, delivery and fast partner onboarding.
Mapping Agent
working codename “Cartos” · pending trademark clearance
Generates and refines field/schema mappings between any two formats.
Drafts HL7 v2 ↔ FHIR ↔ X12 ↔ NCPDP mappings automatically, handles Z-segments and custom formats, and keeps every mapping versioned with rollback and plain-English explanations.
AI technique: LLM + semantic schema matching + retrieval over prior mappings
What it does
- AI-drafted field & segment mappings
- Custom / proprietary Z-segment handling
- Reusable mapping templates
- Versioning, diff & one-click rollback
- Mapping explainability (why each field maps)
Terminology Agent
working codename “Lexid” · pending trademark clearance
Translates and normalizes codes across clinical & billing code systems.
Crosswalks ICD-10, SNOMED CT, LOINC, CPT/HCPCS, RxNorm and NDC, binds value sets to US Core, and reconciles messy local codes to standards — the backbone of medical-coding interoperability.
AI technique: Semantic matching with human-in-the-loop confirmation
What it does
- ICD-10 · SNOMED · LOINC · CPT · RxNorm · NDC crosswalks
- Local-code → standard-code reconciliation
- Value-set & US Core binding
- Confidence scoring on every match
Transformation Agent
working codename “Morphi” · pending trademark clearance
Converts every message in both directions — without losing meaning.
Parses, converts and serializes messages in both directions with round-trip fidelity — real-time, batch, or streaming — including Bulk FHIR export for population-scale jobs.
AI technique: Adaptive parsing of proprietary variants; auto-repair
What it does
- Bidirectional v2 ↔ FHIR conversion
- Round-trip fidelity guarantees
- Real-time, batch & streaming pipelines
- Bulk FHIR ($export / NDJSON)
- Auto-repair of malformed messages
Validation Agent
working codename “Conforma” · pending trademark clearance
Guarantees output conforms to the correct profile / implementation guide.
Validates against US Core, Da Vinci, IPS and CA Core+ profiles, catches schema errors before they ship, and produces the conformance reports procurement and auditors ask for.
AI technique: Explains validation failures and suggests fixes
What it does
- US Core / Da Vinci / IPS / CA Core+ validation
- Schema & cardinality checks
- Procurement & audit conformance reports
- Plain-English fix suggestions
Orchestration Agent
working codename “Fluxa” · pending trademark clearance
Routes and reliably delivers everything.
Connectors, routing rules, error handling with retry/replay, guaranteed delivery and SLA management — the reliable backbone under the fleet.
AI technique: Auto-triages failures; self-healing retries
What it does
- EHR / lab / pharmacy / payer connectors
- Routing rules & pipelines
- Error handling, retry & replay
- Guaranteed delivery & back-pressure
Onboarding Agent
working codename “Initvia” · pending trademark clearance
Stands up a new partner fast.
Learns a partner's format from sample messages and auto-bootstraps the first interfaces — the agent that turns onboarding from weeks into days.
AI technique: Learns partner format from samples; bootstraps interfaces
What it does
- EHR / lab / pharmacy / payer onboarding
- Endpoint discovery
- Sample-driven mapping bootstrap
- Sandbox provisioning
Payer, prior auth & the controls that pass procurement
The agents that ride the CMS-0057-F deadline and unlock payer and health-system sales: end-to-end electronic prior authorization plus consent, provenance and audit.
Prior-Auth Agent
working codename “Authra” · pending trademark clearance
Runs end-to-end electronic prior authorization.
Powers the CMS-0057-F Prior Authorization (PARDD) API and Da Vinci PAS/CRD/DTR, bridges X12 278 ↔ FHIR, assembles documentation automatically and handles denial reasons.
AI technique: Predicts PA requirements; gathers documentation; drafts appeals
What it does
- CMS-0057-F PARDD API
- Da Vinci PAS / CRD / DTR
- X12 278 ↔ FHIR bridging
- Automated documentation assembly
- Denial-reason handling & appeals drafting
Compliance Agent
working codename “Custodia” · pending trademark clearance
Enforces security, consent and provenance.
Consent management, provenance, immutable audit, SMART/OAuth2/OIDC policy and PHI guardrails — the controls that make selling to payers and health systems possible.
AI technique: Detects abnormal access & policy violations
What it does
- Consent management
- Provenance & immutable audit
- SMART on FHIR / OAuth2 / OIDC policy
- RBAC & PHI-handling guardrails
The drug-industry interoperability surface
One agent that owns e-prescribing, EPCS, real-time benefit, electronic prior auth and RxNorm↔NDC normalization — opening the pharmacy, PBM and pharma buyer set.
Pharmacy Agent
working codename “Rxena” · pending trademark clearance
Owns the drug-industry interoperability surface.
Handles e-prescribing (NCPDP SCRIPT), EPCS, electronic prior auth, Real-Time Prescription Benefit, medication history and RxNorm↔NDC normalization — connecting EHRs, pharmacies, PBMs and pharma.
AI technique: Benefit / alternative-drug suggestions; med reconciliation; safety signals
What it does
- e-Prescribing (NewRx, Renewal, Change, CancelRx, RxFill)
- EPCS (controlled substances)
- Real-Time Prescription Benefit (RTPB)
- Electronic prior auth (ePA)
- RxNorm ↔ NDC normalization & med history
Identity, documents and self-healing operations
The agents that turn raw exchange into clean, resolved, observable data: patient identity, document NLP and anomaly-detecting operations.
Identity Agent
working codename “Idenca” · pending trademark clearance
Resolves patient identity and cleans data.
Master patient index / matching, de-duplication and merge, normalization and enrichment — so records line up across every connected system.
AI technique: Probabilistic + ML matching; anomaly detection
What it does
- Patient matching / EMPI
- Dedup & merge
- Normalization & enrichment
- Referential-integrity checks
Document Agent
working codename “Documa” · pending trademark clearance
Moves between clinical documents and structured data.
Converts C-CDA ↔ FHIR, generates and parses documents, and extracts structured data from unstructured clinical text with clinical NLP.
AI technique: Clinical NLP, entity extraction, summarization
What it does
- C-CDA ↔ FHIR
- Document generation & parsing
- Unstructured → structured extraction (NLP)
- Attachments (Da Vinci CDex)
Observability Agent
working codename “Vigila” · pending trademark clearance
Surfaces problems before they become outages.
Monitoring, dashboards, alerting and anomaly detection on message volume and patterns — so problems surface before they become outages.
AI technique: Anomaly detection on message patterns; predictive alerts
What it does
- Real-time monitoring & dashboards
- Throughput / latency SLAs
- Anomaly detection
- Predictive alerting
Agent codenames are working names pending trademark & domain clearance. Functional names describe what each agent does.
Security & Compliance
Enterprise-grade trust, stated honestly
Protecting PHI is non-negotiable. Here's exactly where our certifications stand — no overclaiming.
Compliant · BAAs available
Audit in progress
On roadmap
Native, by design
AWS HIPAA-eligible infrastructure · Business Associate Agreements (BAAs) · Encryption in transit & at rest · See our security details →
Frequently asked questions
Want to see the fleet run on your own message types?
Bring a real HL7 v2 feed or FHIR requirement. We'll show the agents map, translate and validate it live.