Skip to main content
Glama

Server Details

Cardiopulmonary exercise test (CPET) analysis: ventilatory thresholds, metrics and substrate use.

Ownership verified
Status
Healthy
Uptime
100.0% over the last 24 hours
Last Tested
Transport
Streamable HTTP · MCP 2025-11-25
URL
ChatGPT
–
Claude
Oxynet

TDQS

A4.2/5.0

Scored across 13 tools

Disambiguation4/5

Most tools target distinct actions or resources, and descriptions explicitly disambiguate assess_cpet vs the dedicated analyze_* tools vs compute_metrics. Minor overlap remains between upload_cpet and create_upload for disk files, and between get_capabilities and list_metrics for metric information.

Naming Consistency5/5

All tool names use snake_case with a leading verb and domain noun, e.g. analyze_vt, upload_cpet, get_cpet_series, compute_metrics. The pattern is predictable throughout, with no camelCase or mixed-style deviations.

Tool Count5/5

The 13 tools are well-scoped for a CPET analysis API: upload and sample creation, inspection, metrics discovery, full and specialized analyses, and deletion. Each tool covers a meaningful lifecycle stage rather than adding redundant surface.

Completeness4/5

The surface covers upload from both conversation and disk, sample loading, recording inspection, signal slicing, capabilities, metric discovery, full assessment, specialized analyses, and deletion. Minor gaps include no tool to list uploaded handles or re-fetch prior analysis results, so lost results require a costly re-run.

Publisher details

Operator
Oxynet · Publisher source
Operator website
https://www.oxynet.net/
Vendor relationship
Not available
Documentation
Not available
Privacy policy
Not provided
Trust center
Not available
Restrictions
Not available

Available Tools

13 tools
analyze_eovAnalyse oscillatory ventilation (beta)AInspect

Analyses exercise oscillatory ventilation and the ventilatory control loop behind it in an uploaded recording. Beta: developed on one heart failure cohort, transportability untested. Needs end-tidal CO2; without it the analysis is refused with that reason.

Returns cpet_id and one envelope in results: status ("ok" or "not_analysable"), findings, quality (good/acceptable/poor, of the INPUT, not of the answer), confidence ("low" or "normal" with the reasons, never a number), notes, provenance, and error/message when it could not run. Counts one unit of the key's monthly allowance, a repeated call included. Licensed separately (get_capabilities lists this key's analyses).

The findings answer three different questions:

  • GRADE and BURDEN: how MUCH oscillation there was.

  • MARGIN.LOOP_GAIN: how CLOSE the control loop is to oscillating on its own, 0 stable and 1 self-sustaining. Defined even when the grade is "none": a grade saturates and this does not.

  • LOOP: the factorisation, when an oscillation is clear enough to invert. g = plant_gain x controller_gain. delay_s comes from the phase and is NOT a factor of g.

And three limits:

  • margin.corroborated false: end-tidal CO2 did not oppose ventilation at that period, so what resonates is not the chemoreflex loop. It is ventilatory variability, not instability.

  • delay_s is the whole lag in the return path, not a circulation time. "Circulation" would be an interpretation, not a measurement.

  • co2_amplitude_suspect: the end-tidal CO2 waveform was distorted (a clipped sample) while ventilation stayed clean. Plant gain is then a LOWER bound and controller gain an UPPER one.

An absent note does not mean an absent finding.

Inputs: cpet_id (required).

analyze_substrateAnalyse fat and carbohydrate useAInspect

Analyses fat and carbohydrate oxidation from gas exchange in an uploaded recording: FATMAX, peak fat oxidation, the crossover to carbohydrate and gross efficiency, with stage means and fitted curves. For the headline FATMAX numbers alone, compute_metrics with ["substrate_summary"] costs about a sixth of the tokens.

Returns cpet_id and one envelope in results: status ("ok" or "not_analysable"), findings, quality (good/acceptable/poor, of the INPUT, not of the answer), confidence ("low" or "normal" with the reasons, never a number), notes, provenance, and error/message when it could not run. Counts one unit of the key's monthly allowance, a repeated call included. Licensed separately (get_capabilities lists this key's analyses).

Inputs: cpet_id (required), curves.

analyze_vtFind ventilatory thresholdsAInspect

Finds VT1 and VT2 in an uploaded recording with Oxynet's models trained on expert-labelled tests: the time of each and the VO2 at it. Unnamed, the model best suited to the recording answers, named with the reason in provenance.selection. Everything at once is assess_cpet; a single number is compute_metrics.

Returns cpet_id and one envelope in results: status ("ok" or "not_analysable"), findings, quality (good/acceptable/poor, of the INPUT, not of the answer), confidence ("low" or "normal" with the reasons, never a number), notes, provenance, and error/message when it could not run. Counts one unit of the key's monthly allowance, a repeated call included. Licensed separately (get_capabilities lists this key's analyses).

Inputs: cpet_id (required), model_name.

assess_cpetAssess a recordingAInspect

One call on an uploaded recording: the kind of test, VT1 and VT2 from the best-suited method, oscillatory ventilation (beta: one heart-failure cohort, transportability untested) and substrate use when the channels allow, and the derived metrics. Runs only what the key includes; plan says why a part did not. confidence per result, repeated in headline. One unit per analysis run, a repeated call included.

Inputs: cpet_id (required).

compute_metricsCompute derived metricsAInspect

Computes derived quantities on an uploaded recording: cheaper than a full analysis when a number is all that is needed. Each call counts one unit of the key's monthly allowance, a repeated call included.

Each result is {"status": "ok", "value": ...} or {"status": "unavailable", "error": ..., "message": ...}. A metric whose requirements are unmet is refused with the reason, never computed on inadequate input and never returned as a bare null.

  • substrate_summary: the headline FATMAX numbers for about a sixth of the full substrate analysis. fatmax_found false means fat oxidation was still climbing at one end of the range, so FATMAX lies outside what the protocol covered (a fact about the protocol, not a missing value). steady_state false means the test was a ramp: FATMAX reads high and is not comparable with a graded test. windows_above_rer_one counts the windows where the fat number is an upper bound, because above RER 1.0 exhaled CO2 carries a non-metabolic part.

  • ve_vco2_slope: a profile over the first 25/50/75/100 % of exercise, not one number, because a slope that climbs across a test is a different finding from a flat one.

  • phases: rest, warm-up, exercise and recovery, inferred from the signals with an error of tens of seconds, so no boundary is exact. An absent rest phase is an answer, not a failure.

  • gas_quality and sampling_adequacy: whether the recording can support a measurement at all.

Inputs: cpet_id (required), metrics (required).

create_uploadCreate upload linksAInspect

One-time URLs for uploading CPET files that are on disk, named by a path.

Returns an upload_url and a ready-to-run curl line. POSTing the file to it (multipart field file) returns the same summary as upload_cpet, cpet_id included. The bytes never enter the conversation, and the URL carries its own five-minute credential, so the command handles no API key.

curl -sS -F "file=@/path/to/export.csv" "<upload_url>"

For a folder, count mints every URL in one call (upload_urls), so one shell command can POST every file: a cohort takes two steps in total.

files=(/data/*.xls); urls=(URL1 URL2 ...)
for i in "${!files[@]}"; do
  curl -sS -F "file=@${files[$i]}" "${urls[$i]}"; done

Each URL accepts exactly one file, and they all expire five minutes after minting.

Inputs: count, filename, retain_hours.

delete_cpetDelete a recordingA
DestructiveIdempotent
Inspect

Delete an uploaded recording immediately, instead of at its 24 h expiry.

For when the user has finished with a recording, particularly clinical data. It cannot be undone: the uploaded file was discarded at parsing, so analysing the recording again takes a new upload.

Inputs: cpet_id (required).

get_capabilitiesCheck what this key may doA
Read-onlyIdempotent
Inspect

Reports what this API key may do, without running anything: the analyses it is licensed for (analyze_vt, analyze_eov, analyze_substrate), the threshold models it can reach, the metrics compute_metrics may compute for it, its upload limits and the data-retention policy.

Meant as the first call of a session. The products are licensed separately, so a key with thresholds does not necessarily include oscillation analysis, and this answers that before any analysis is tried. Every operation it lists is run by its own tool, never by this one.

Inputs: no named parameters.

get_cpetSummary of a recordingA
Read-onlyIdempotent
Inspect

What is actually in an uploaded recording, without pulling the signals.

The signal-processing view of a file: the detected vendor format, which channels are present and how complete each one is, the sampling interval the FILE arrived at (before Oxynet resampled it), the external load and the protocol reconstructed from it, and every parser flag raised on the way in. The same summary upload_cpet returned, re-read on demand, so a conversation that has lost it needs no second upload. An unknown or expired handle returns CPET_NOT_FOUND, here and in every tool that takes one.

  • A channel with present: false carries a reason. Absent and present-but-reading-zero are different facts, and the reason says which.

  • load.available: false is an answer, not a gap. Most recordings carry no ergometer channel, and nothing is back-calculated from VO2 to invent one, so analyses that need watts report that they cannot run.

flags describes the RECORDING. It is not a confidence in any result.

Inputs: cpet_id (required).

get_cpet_seriesSeries for a chartA
Read-onlyIdempotent
Inspect

A downsampled slice of the stored signals, for plotting the data or looking at a specific stretch of a recording. Not for analysis.

Every measurement Oxynet offers runs server-side on the full record at full resolution. This slice is decimated by stride, so a narrow feature can fall between two kept samples, and arithmetic on it is costlier and less accurate than the corresponding metric.

Inputs: cpet_id (required), channels, max_points.

get_sample_cpetLoad the synthetic sampleAInspect

Load Oxynet's synthetic CPET (not a real person) and get a cpet_id back, ready to analyse: for demonstrating Oxynet or testing a connection without a patient file. Real CPET exports are health data, and their filenames often carry patient names.

Returns the same summary upload_cpet returns. The recording is analysed like any other and deleted after 24 h, or at once with delete_cpet.

The thresholds in the plain sample sit near 55 % and 80 % of the exercise period because they were placed there. It is a demonstration fixture, not validation data and not evidence of accuracy.

Inputs: oscillating.

list_metricsList derived metricsA
Read-onlyIdempotent
Inspect

Every derived quantity Oxynet can compute, with what each one needs.

The registry covers peak values (VO2max, VEmax, RERmax, O2 pulse), the VE/VCO2 slope profile, substrate use and FATMAX, and input-quality checks (gas-channel agreement, sampling adequacy, clock integrity).

Each entry states its unit, the channels it requires, and the traps in reading it. It is the list of valid names for compute_metrics, and it grows without this tool changing.

Inputs: no named parameters.

upload_cpetUpload a CPET recordingAInspect

Upload a CPET export whose contents are already in the conversation (pasted or attached) and get a cpet_id back.

The file is sent exactly as exported, unchanged: Oxynet does the parsing. It detects the vendor format and knows its unit conventions, per-phase clock resets and decimal-separator defects. A file parsed or converted beforehand loses that (a VO2 column in L/min read as mL/min is off by 1000, and nothing downstream notices).

For a file on disk, a call with no arguments returns the same one-time upload link as create_upload, so the bytes go from disk to Oxynet without entering the conversation (a 360 KB export is about 97,000 tokens as text).

UNSUPPORTED_FORMAT means no parser recognised the file. Its recovery block gives the canonical JSON shape, the required channels and units, and the formats considered, for a file converted into that shape.

Returns cpet_id (the handle every other tool takes), the detected source format, sampling as the file arrived, channels with their coverage, and flags. The recording stays on the server.

Privacy: the uploaded bytes are parsed and discarded, never stored, and the parsed record is deleted after 24 h. Oxynet needs no patient identifiers; CPET exports often carry the patient's name in the filename and header.

Inputs: file, content, filename, retain_hours.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      Not graded
      maintenance
      Enables computational CPET analysis by calculating VE/VCO2 ventilatory equivalent slope, pulmonary dead space fraction, V-slope anaerobic threshold, and a cardiorespiratory efficiency index to support Zone 2 and threshold pacing directives.
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables deterministic calculation of HCVR slope, central respiratory chemosensitivity, and CO2 tolerance from end-tidal CO2 and spirometry inputs. It also provides chemoreflex loop-gain stratification for personal-agent actions.
      7
      MIT
    Try in Browser

    Glama MCP Gateway

    Add one secure layer between your agents and this server.

    Resources