Developers and partners

Build on a computational layer for CPET

Oxynet turns CPET recordings into structured, reproducible physiological measurements that people, clinical software, research workflows and AI agents can consume. Send a supported recording, receive measurements with their provenance and the quality of the input they came from.

Quick start →Downloads Test one of your files

API v1 · docs 1.0 · updated 2026-09-22

Where Oxynet sits

Oxynet is not an application a laboratory installs, and not a single threshold algorithm. It is the computation underneath: it reads the file the metabolimeter exported (23 format ids, detected automatically), measures the physiology in it, and hands structured results to whatever sits downstream.

Acquisition hardware, manufacturer software and clinical workflows stay as they are. What changes is that interpretation becomes a call, made the same way every time, with every result traceable to the model and version that produced it.

VT1 and VT2 are the entry point, not the ceiling: substrate use, derived quantities, signal integrity and, in research, oscillatory ventilation come back through the same interface. The analysis catalogue states the evidence behind each one.

CPET device / laboratory

CPET recording

vendor export, unchanged

Oxynet

parse, condition, measure

Structured measurements

JSON with provenance

Clinical software, research, reports, AI agents

Four ways in

REST API

Production

Clinical platforms, laboratory systems, automated pipelines.

Upload the vendor export, work with the handle, receive JSON. One key, analyses granted per key.

Python

Available

Researchers, analysts, scientific workflows.

Call the hosted engine from Python over plain HTTP. The open-source pyoxynet package covers local research inference separately.

MCP

Available

AI agents and agentic workflows.

A native connector for Claude, Gemini CLI and any MCP client. The agent orchestrates and explains; Oxynet computes the physiology.

Local / embedded

By arrangement

Partners needing on-premise processing or product integration.

The same engine on your infrastructure or in your product. Scoped per partner, not self-service.

What is live, and what is not

Capabilities do not all stand on the same ground, and the difference matters before you build on one. Three states are used consistently here and in the manual:

Production

Answering on the live API, with the evaluation that let each model ship stored beside it.

Available

Shipped and usable: the interfaces and deployment options.

Research

Running, and not yet shown to transport to a new population.

REST API (v1)	Production	Key per caller, analyses granted per key.
MCP endpoint	Available	Streamable HTTP at /oxynet-mcp, for any MCP client.
OpenAPI 3.0 schema	Available	Public, for client generation and agent actions.
Web application	Available	app.oxynet.net, for results without building anything.
Python package (pyoxynet)	Available	Open source, local research inference. Not a client for the API.
Local and embedded deployment	By arrangement	Scoped per partner. Not a self-service product.

Three things to know before the first call

Hand over the file, keep the handle

Upload the export unchanged. Converting units yourself is where errors of a thousandfold come from. Every later call names the returned cpet_id.

No confidence scores

Nothing is calibrated against clinical outcomes. quality describes the recording, never the certainty of the answer.

A refusal is a result

When a recording cannot support an analysis, the API says so and says why. Report it; do not work around it.

Regulatory status

Oxynet is research software. It is not a medical device, carries no CE mark or FDA clearance, and its output is not a diagnosis. Interpreting a measurement for a patient remains the clinician's act.