What does the Fonteum platform actually expose?

One source-provenanced US healthcare provider graph, drawn from 44 federal source families and reachable four ways: a REST API, an MCP server for AI agents, an HL7 FHIR R4 interface, and bulk dataset export. Every path returns the same provenance on every field.

What is the 14-tuple provenance contract?

Each rendered value carries 14 provenance fields — dataset ID, source agency, source URL, archive URL, snapshot date, reporting period, methodology version, entity type, entity ID, field name, field value, confidence tier, chain-link hash, and attested-at timestamp. The chain ties every fact to the federal row it came from.

How do point-in-time snapshots and bitemporal reads work?

Every record carries two time axes: when a fact was true and when Fonteum first knew it. A point-in-time read returns what a record said on a past date, and what was known about it then. History is accrued at ingest, never reconstructed after the fact.

Can an AI agent call the platform directly?

Yes. The MCP server lets an agent resolve a provider, screen against exclusion lists, and pull provenance without a human in the loop. The same data is also published as machine-readable dataset declarations and an agent card for autonomous discovery.

Is the platform data free to read?

Reads of public federal and state records are free and carry no paywall, so AI search engines can cite them. Authentication raises rate limits and unlocks bulk export and a signed attestation artifact for compliance workflows.

What proves the data has not been altered after the fact?

Each snapshot is signed and chained: the raw source bytes are hashed at capture, the hash is linked into an Ed25519 witness chain, and the chain can be re-derived on demand. The attestation pack turns that integrity into a record an auditor can check independently.

PLATFORM · THE CAPABILITY LAYER

Every capability behind the provenance graph.

Fonteum is one source-provenanced US healthcare provider graph drawn from 44 federal source families. The platform exposes it as a REST and MCP API, FHIR R4, bulk export, point-in-time snapshots, an entity graph, and signed attestation — every field carrying a 14-tuple provenance contract.

Request access Read the API reference →

The capabilities

Eight capabilities, one graph

Query layer

REST API

Resolve a provider across NPPES, OIG LEIE, CMS PECOS, Care Compare, and Open Payments in one request — every field carrying its 14-tuple provenance.

Open REST API→

Agent layer

MCP server

A Model Context Protocol server an AI agent can call directly. Same provider graph, same provenance, structured for LLM consumption.

Open MCP server→

Interop

FHIR R4 API

Five US Core 6.1.0 provider resources — Practitioner, Organization, Location, PractitionerRole, HealthcareService — with row-level provenance on every field.

Open FHIR R4 API→

Egress

Bulk export

Whole-dataset downloads as gz-compressed NDJSON and CSV across every ingested federal source family, with a machine-readable manifest.

Open Bulk export→

Evidence

Attestation & audit pack

A signed, chained attestation wrapping any check — the timestamped, source-cited record a compliance or procurement audit asks for.

Open Attestation & audit pack→

Cross-source

Reconciliation

Source-versus-source diffs that show where two federal records disagree on the same entity, and which snapshot each assertion came from.

Open Reconciliation→

Identity

Entity graph

Providers, organizations, and facilities resolved across NPI, CCN, PECOS-ID, and chain identifiers — deterministic links at full confidence.

Open Entity graph→

Time-travel

Snapshots

Point-in-time, bitemporal reads. Ask what a record said on a past date and what was known about it then — the history is accrued, not back-filled.

Open Snapshots→

How it fits together

One pipeline, from federal record to callable fact

The capabilities above are stages of a single pipeline. Each fact stays tied to the source row it came from at every stage.

Ingest — federal and state public records are pulled on a per-source cadence; raw bytes are hashed and retained.
Reconcile — records are resolved into a single entity graph across NPI, CCN, PECOS-ID, and chain identifiers.
Attest — every value is chained to a 14-tuple provenance contract and signed against an Ed25519 witness chain.
Serve — the same graph is exposed over REST, MCP, FHIR R4, and bulk export, with point-in-time reads on every record.

Platform FAQ

Common questions

What does the Fonteum platform actually expose?: One source-provenanced US healthcare provider graph, drawn from 44 federal source families and reachable four ways: a REST API, an MCP server for AI agents, an HL7 FHIR R4 interface, and bulk dataset export. Every path returns the same provenance on every field.
What is the 14-tuple provenance contract?: Each rendered value carries 14 provenance fields — dataset ID, source agency, source URL, archive URL, snapshot date, reporting period, methodology version, entity type, entity ID, field name, field value, confidence tier, chain-link hash, and attested-at timestamp. The chain ties every fact to the federal row it came from.
How do point-in-time snapshots and bitemporal reads work?: Every record carries two time axes: when a fact was true and when Fonteum first knew it. A point-in-time read returns what a record said on a past date, and what was known about it then. History is accrued at ingest, never reconstructed after the fact.
Can an AI agent call the platform directly?: Yes. The MCP server lets an agent resolve a provider, screen against exclusion lists, and pull provenance without a human in the loop. The same data is also published as machine-readable dataset declarations and an agent card for autonomous discovery.
Is the platform data free to read?: Reads of public federal and state records are free and carry no paywall, so AI search engines can cite them. Authentication raises rate limits and unlocks bulk export and a signed attestation artifact for compliance workflows.
What proves the data has not been altered after the fact?: Each snapshot is signed and chained: the raw source bytes are hashed at capture, the hash is linked into an Ed25519 witness chain, and the chain can be re-derived on demand. The attestation pack turns that integrity into a record an auditor can check independently.

Go deeper

Where to go next

BUILD ON THE PLATFORM

Resolve, screen, and attest on public federal data.

Start against the REST API or drop the MCP server into an agent. Request access for higher limits, bulk export, and signed attestation artifacts.

Add Fonteum to your agent →Request pilot access