Healthcare & Operations · flagship product

The first GenAI-native healthcare integration engine.

OpenLI HIE turns natural language into the development language for healthcare integration. Engineers describe an HL7 mapping, a FHIR transformation, or an IRIS production change in plain English — the runtime designs, builds, tests, deploys, monitors and debugs it, under human approval, with audit trails on every action and one-command rollback.

Built on the OpenLI Codex foundation. Used by NHS Trust integration teams who need to ship faster without losing control of their HL7 v2, FHIR R4, IRIS, file and database transports.

v1.9.7 production Multi-tenant 7-tier RBAC AGPL-3.0 / commercial dual Whitelabel-ready
5Microservices in the HIE stack
4Transport types: HL7 v2, FHIR R4, REST, file, db
7RBAC role tiers, audit-logged on every change
0Restarts required for hot config reload

Natural language is the development language.

Healthcare integration is one of the most expensive, slowest and most clinical-safety-critical parts of any NHS digital programme. HIE was designed to compress that lifecycle radically — without compromising the governance NHS teams cannot do without.

Plain English, end-to-end

From requirement to deployed integration, the developer interface is natural language. Engineers describe what they need; the runtime drafts the mapping, the production, the route, the test fixture — and presents it for review.

Clinical safety guardrails

Pre-deployment policy hooks sanitise PII before any data is presented to the model. RBAC controls who can change what. Approval workflows gate every state transition. Audit trails record every agent action.

Same governance as Codex

HIE inherits the OpenLI Codex governance model. One security review covers HIE plus every other OpenLI product your Trust adopts. One audit story. One operational model.

Capabilities (v1.9.7, in production today)

The capabilities below ship in HIE today. They are the actual product, not a roadmap.

HL7 v2 transport

Bidirectional HL7 v2 interfaces with version-aware parsing, segment validation, and natural-language mapping authoring. Handles ADT, ORM, ORU, SIU, MFN and the long tail of NHS message types.

FHIR R4 transport

FHIR R4 RESTful endpoints with resource validation, profile conformance and natural-language transformation between v2 and FHIR. Supports both producing and consuming FHIR APIs.

HTTP/REST & file & database

Generic HTTP/REST adapter, file watcher with rotation handling, and direct database read/write transports. The same natural-language tooling applies across all transport types.

Hot reload

Configuration changes apply without restarting the engine. No dropped messages, no clinical-system disruption, no scheduled outage windows for routine configuration changes.

Config snapshots & rollback

Every configuration change creates an immutable snapshot. One-command rollback to any previous snapshot if a deployment misbehaves. Rollback never destroys forensic evidence.

Multi-tenant workspace isolation

Each NHS Trust gets its own isolated workspace with its own configuration, audit log, RBAC policy and tenant branding. One HIE installation can serve a managed-service operator hosting multiple Trusts.

5-microservice architecture

Portal (operator UI) · Agent Runner (Codex-powered agentic runtime) · Manager API (state and policy) · Engine (the integration runtime itself) · Prompt Manager (versioned skills and templates).

7-tier RBAC

Granular role-based access control aligned with NHS operating models: super_admin, org_admin, project_admin, integration_engineer, clinical_reviewer, viewer, plus tenant-scoped variants. All actions audit-logged.

Open source core

AGPL-3.0 with a commercial dual licence for enterprises. Auditable by your security team. Self-hostable behind the NHS firewall. The core is at github.com/zhongli1990/hie.

How it works

From requirement to clinically-safe deployed change, in plain English, in five steps.

1. Describe

An engineer (or a clinical analyst) describes the change they need in plain English. "Add a new HL7 v2 ORM route from PACS to the EPR, mapping the accession number to OBR-3."

2. Draft

The Codex-powered agent runtime drafts the configuration change, the mapping, the test fixture, and the rollback plan — presented as a reviewable diff in the operator portal.

3. Review

The clinical reviewer (or integration lead) approves, rejects or comments on the draft. Required approvers are defined per workspace and per change type. Nothing deploys without sign-off.

4. Deploy

On approval, the change applies via hot reload — no engine restart, no dropped messages. A snapshot is created automatically for rollback. The audit trail records every step.

5. Monitor & rollback if needed

Live message throughput, error rates and per-route counters are visible in the portal. If something misbehaves, one-command rollback restores the previous snapshot in seconds.

Defined integration use cases

HIE ships with prompt templates and skills for the use cases below. Each is a versioned, reviewable artefact in the prompt manager — not magic strings hidden in code.

HL7 v2 mapping authoring

Author a new mapping from a natural-language description; produce a reviewable mapping spec plus test fixtures.

FHIR transformation

Convert HL7 v2 messages to FHIR R4 resources with profile-aware validation. Handle the edge cases that the standard FHIR tooling does not.

Production change planning

Plan a multi-step production change with explicit stages, risk assessment and rollback plan, all generated from a plain-English requirement.

Test fixture generation

Generate realistic synthetic HL7 v2 / FHIR test fixtures that exercise the new mapping, including edge cases the engineer may not have thought of.

Incident triage

When a route starts erroring, the agent runtime reads the recent logs, the route configuration, and the recent changes — and proposes a hypothesis with evidence.

Documentation generation

Generate up-to-date interface specifications, integration runbooks and clinical-safety case files from the live configuration. Documentation never goes stale.

Built on OpenLI Codex

HIE is not a standalone product. It is built on the same agentic runtime that powers every other OpenLI product. The clinical safety review you do for HIE covers Integrai, IRIS CoPilot, IMX Monitor and DMM — one security posture, one operational model.

FOUNDATION

OpenLI Codex powers HIE

The dual-runner architecture (Claude Agent SDK + OpenAI Codex SDK), the 7-tier RBAC, the audit trail, the prompt template library and the workspace registry — HIE inherits all of these from the Codex foundation.

Whitelabel for managed-service operators

HIE is whitelabel-ready for system integrators and managed-service operators who run integration stacks for multiple NHS Trusts.

Multi-tenant by design

Workspace isolation per Trust. Each Trust’s configuration, audit log and operator UI is fully separate from the others. One HIE deployment serves many Trusts cleanly.

Per-tenant branding

Logo, colour, copyright notice and portal domain are per-tenant. Trust users see their own brand, not yours.

OEM commercial model

OEM licence or revenue share — we work with the model that fits your channel. Talk to the team to scope a partnership.

HEALTHCARE INTEGRATION

Ready to ship integrations in plain English?

Open the HIE portal, view the source on GitHub, or talk to the team about deploying HIE for your NHS Trust or managed-service operation.