Construe API

This API designed to bridge the gap between free-text clinical documentation and the structured data formats that health systems, payers, and analytics platforms need. It applies advanced clinical natural language processing (cNLP) and knowledge-graph–driven interpretation to extract and normalize key medical concepts from any unstructured source.

Inputs: Almost any raw clinical text from sources like EHR progress notes, radiology reports, discharge summaries, or patient intake forms. Data can be sent in plain text, JSON, or as part of an HL7/FHIR payload.

Data processing: Construe uses domain-trained language models to understand clinical context, disambiguate terms, and handle shorthand, typos, and varying documentation styles.

Outputs: Key data points — including conditions, medications, procedures, and lab results — are identified, normalized, and enriched with standardized codes. Choose between fully structured JSON or a hybrid output that preserves original text snippets alongside structured codes. All processing is HIPAA-compliant and performed on secure infrastructure.

Supported vocabularies: CPT, ICD-10-CM, ICD-10-PCS, SNOMED, LOINC, RxNorm, HPO, and custom.

Lang2FHIR API

This API is designed to take the messy, unstructured reality of healthcare language and turn it into clean, standards-compliant FHIR® resources you can use immediately in your systems.

Inputs: Almost any raw clinical text.

Data processing: Lang2FHIR is trained on both general clinical language and specialty-specific vocabularies. It recognizes not only terms but also context — negations (“no history of diabetes”), temporality (“started 2 weeks ago”), and relationships between entities.

Outputs: Converts extracted information into a properly structured FHIR resource. The API returns a JSON bundle of fully-formed FHIR resources, ready to store in your EHR, feed into analytics pipelines, or trigger alerts in clinical decision support systems.

Workflows API

Getting from point A to point B in healthcare AI development is never a straight line. With Workflows, all you have to do is describe what you want to happen, and Workflows will fill in the blanks.

Example prompts

• “When my app loads, I want to a secure patient message translated into FHIR resources and sent to Medplum.”
• “When a new observation of the following type is created, create a corresponding condition.”

Output: A detailed graph of the exact steps to take with detailed instructions on how to use Phenoml APIs to connect all the dots.

Agents API

Your AI-native healthcare assistant. A single API that coordinates multiple specialized models to handle complex clinical and operational requests.

Inputs: Send any healthcare-related question or task prompt — from “summarize this patient’s chart” to “draft a prior authorization letter” — to the Agent API.

Data processing: Automatically routes the request through the right combination of domain-specific AI tools, including Lang2FHIR, Construe, and custom reasoning modules. It pulls in relevant patient data, clinical guidelines, and terminology standards to ensure the response is accurate, compliant, and medically grounded.

Outputs: Returns a ready-to-use result — whether that’s structured data, a written document, or a decision-support recommendation — optimized for direct use in workflows or EHRs.