Anpheros is a FHIR R4 healthcare API where each patient controls their own record, with a free sandbox, SDKs and an MCP server for AI agents.
One patient-controlled HL7 FHIR R4 record per person · an API for applications, clinics, laboratories and AI agents · consent, provenance and an audit built in.
Developer guides · API reference · SDKs & examples · Platform · anpheros.com
Healthcare API · Healthcare API platform · FHIR platform
Anpheros is the medical data layer a healthcare product is built on. Each person has one record made of standard HL7 FHIR R4 resources — not a proprietary schema — and applications read and write it through an API:
- for patients it is a digital medical record they control (the Anpheros Daily app);
- for developers it is a managed FHIR data store — a FHIR backend — with patient consent, provenance on every write, an access log the patient can see, and an AI context API that prepares the relevant part of a record for the model of your choice.
Anpheros' own patient app is a client of the same public API as everyone else, so what it records is available — with the patient's consent — to other applications in the same FHIR form.
Your application · clinic system · lab system · AI assistant or agent
│ HTTPS — API key (your patients) or OAuth token (one consented patient)
▼
Anpheros API /fhir/R4 (strict FHIR) · /v1 (simplified REST, same data and ids) · /oauth (consent)
│ isolation per application · scopes · provenance · access log
▼
One HL7 FHIR R4 record per patient — every version kept — shared only with the patient's consent
| Instead of building… | …you use |
|---|---|
| a medical data model and tables | 26 FHIR R4 resource types: observations, conditions, medications, documents, immunizations, allergies, encounters, care plans and more |
| export formats for every partner | FHIR R4 in and out, International Patient Summary documents, SMART on FHIR for third-party apps |
| a consent system | OAuth 2.1 + PKCE, SMART v2 patient scopes, grants of 30–365 days that the patient can revoke at any time |
| an audit trail | provenance on every write (patient, practitioner, device, import or AI) and every read visible to the patient |
| lab integrations | a lab connector for CSV, HL7 v2 ORU^R01 and JSON reports, coded to LOINC |
| retrieval code for an LLM | an AI context API: the relevant sections of a record within a token budget, every item labelled with its source |
| polling and sync jobs | signed, retried webhooks |
| client libraries | SDKs for TypeScript and Dart / Flutter, with idempotent writes, retries, token refresh and the consent flow |
Get a sandbox key instantly: sign in with Google on the dashboard and press Get a sandbox key. The sandbox is free, and every sandbox project comes with its own 30 synthetic patients.
npm install @anpheros/sdk # or: dart pub add anpheros_sdk
import { Anpheros, apiKey } from '@anpheros/sdk';
const anpheros = new Anpheros({ auth: apiKey(process.env.ANPHEROS_KEY!) }); // sandbox key: sk_test_…
const patient = await anpheros.patients.create({ given: 'Ana', family: 'Example' });
await anpheros.observations.create(patient.id, {
code: '85354-9', display: 'Blood pressure panel', category: 'vital-signs',
effective_at: new Date().toISOString(), author_type: 'device',
components: [{ code: '8480-6', value: 128, unit: 'mm[Hg]' }, { code: '8462-4', value: 82, unit: 'mm[Hg]' }],
});
const ctx = await anpheros.context.build({ patient: patient.id, task: 'weekly check-in', budget_tokens: 1500, format: 'text' });
Runnable examples — a quickstart in TypeScript and Dart, a consent web app, an LLM example and a FHIR transaction — are in anpheros-sdk/examples.
Connect Claude, Cursor, VS Code or any MCP client to FHIR health data: https://platform.anpheros.com/mcp (streamable HTTP), listed in the official MCP Registry as com.anpheros/fhir. With a sandbox key the agent works on your 30 synthetic patients: patient context, International Patient Summary, timeline, FHIR search, lab import and validation. People can also connect their own record through OAuth, with a consent screen and read-only access by default.
claude mcp add --transport http anpheros-fhir https://platform.anpheros.com/mcp --header "X-API-Key: sk_test_…"
Configs for every client: anpheros/mcp-server · guide: MCP server
Anpheros is not an AI platform and does not run models. It is the data layer an AI application or agent calls:
- consented access — an agent reaches one patient's record only through an OAuth grant, within the scopes the patient chose;
- context, not raw dumps —
POST /v1/contextreturns the relevant sections within a token budget, labelled with author type and source, and lists what was left out; - facts kept apart from AI output — values written by an AI are recorded as AI-authored and kept in a separate section of later contexts;
- accountability — every context request is recorded and visible to the patient.
It works with any model — hosted or local — because the model call is your own code. Guides: AI healthcare applications · AI agents and medical data · LLM applications and healthcare data · Building with AI coding tools
| Standard | In Anpheros |
|---|---|
| HL7 FHIR R4 (4.0.1) | native storage and API; CapabilityStatement at /fhir/R4/metadata |
| International Patient Summary | Patient/$summary, validated with the official HL7 validator against the IPS implementation guide (0 errors) |
| SMART App Launch | standalone launch with PKCE and SMART v2 scopes; OpenID Connect fhirUser |
| HL7 v2 | ORU^R01 laboratory results imported as FHIR Observations |
| SMART Health Links | share a record by link or QR code (/shl) |
| Terminologies | LOINC, ICD-10, ATC, UCUM |
EU data residency · isolation between applications with per-application patient identifiers · consent chosen and revocable by the patient · provenance on every write · an access log the patient can see · documents stored immutable and encrypted with a customer-managed key · a separate sandbox with synthetic patients. Details: Security, privacy and data residency.
| Repository | What it contains |
|---|---|
| anpheros-sdk | the official TypeScript and Dart / Flutter SDKs and runnable examples |
| mcp-server | the hosted FHIR MCP server for AI agents: tools and client configs |
| agent-skills | Agent Skill, Claude Code plugin, Cursor rules and AGENTS.md |
| fhir-api-openapi | OpenAPI 3.1 spec and Postman collection |
| fhir-synthetic-patients | 30 synthetic patients as FHIR R4 Bundles and NDJSON |
| ips-examples | International Patient Summary documents in FHIR R4 |
| fhir-validate-action | GitHub Action: validate FHIR R4 JSON in CI |
| lab-results-to-fhir | CLI: CSV / HL7 v2 lab reports → FHIR Observations |
| healthcare-app-starters | Next.js and Flutter health app templates |
All Apache-2.0.
Anpheros Platform is in private beta: EU-hosted, single zone without automatic failover, no contractual SLA yet. The sandbox is free and self-service: sign in with Google on the dashboard and get a key instantly, with 30 synthetic patients per sandbox project. Production is for verified organisations with a data processing agreement, and the first month of production is free. Not available yet: SMART EHR launch.
Anpheros Daily is the patient and family app built on the same record — symptoms, vital signs, medications, lab results, documents, appointments and an assistant that answers with the person's own data. Android, macOS and the web: anpheros.com.
General: contact@anpheros.com · Support: support@anpheros.com · Privacy: privacy@anpheros.com · Security reports: see our security policy.
For AI assistants and search: anpheros.com/llms.txt · developers.anpheros.com/llms.txt · llms-full.txt