generated: '2026-09-14'
method: derived
source: https://aicure.github.io/open_dbm/extras/odbm-rest-api
name: AiCure MCP candidate
status: candidate
summary: >-
  AiCure ships no MCP server. This is a DERIVED candidate tool list showing what an MCP server over
  the published OpenDBM REST API would expose; it is not an agent surface AiCure operates, and no
  MCPServer pointer is emitted in apis.yml. The pointer type used is X-MCPServerCandidate.
deployment:
  mode: none
  endpoint: null
  install: null
  package: null
  auth: unknown
  verified: derived
search_record:
  searched:
  - target: github.com/AiCure - all 41 public repositories
    result: no MCP server; the only first-party projects are open_dbm, open_dbm-Gen1, external_cdx
      and a study-analysis notebook. The rest are forks of GitHub Actions and mobile libraries.
  - target: npm registry search for aicure / opendbm
    result: no package
  - target: PyPI
    result: opendbm only, a measurement library with no MCP transport
  - target: https://api.aicure.com/mcp
    result: HTTP 403 Missing Authentication Token - the AWS API Gateway default for an unmatched
      route, not an MCP endpoint
  - target: /.well-known/* on aicure.com, www.aicure.com, api.aicure.com, aicure.github.io
    result: nothing served; see well-known/aicure-well-known.yml
candidate_tools:
- name: opendbm_login
  rest: POST /odbm/v1/login
  description: Exchange a username and password for a JWT bearer token.
  inputs:
  - username
  - password
- name: opendbm_upload_media
  rest: POST /odbm/v1/upload
  description: Upload a video or audio file to the OpenDBM host and get back its stored path.
  inputs:
  - file
  - file_name
  - file_extension
  - platform
- name: opendbm_video_facial
  rest: POST /odbm/v1/video/facial
  description: >-
    Compute facial biomarkers - facial activity, action units, emotional and pain expressivity, eye
    blink and gaze, facial asymmetry and tremor - from an uploaded video.
  inputs:
  - file_url
  - platform
  - variables
- name: opendbm_video_acoustic
  rest: POST /odbm/v1/video/acoustic
  description: Compute verbal-acoustic biomarkers from the audio track of an uploaded video.
  inputs:
  - file_url
  - platform
  - variables
- name: opendbm_video_movement
  rest: POST /odbm/v1/video/movement
  description: Compute head and upper-body movement biomarkers from an uploaded video.
  inputs:
  - file_url
  - platform
  - variables
- name: opendbm_video_speech
  rest: POST /odbm/v1/video/speech
  description: >-
    Compute speech biomarkers - rate of speech, pause characteristics, lexical richness, parts of
    speech, sentiment - from an uploaded video.
  inputs:
  - file_url
  - platform
  - variables
- name: opendbm_audio_acoustic
  rest: POST /odbm/v1/audio/acoustic
  description: Compute verbal-acoustic biomarkers from an uploaded audio file.
  inputs:
  - file_url
  - platform
  - variables
- name: opendbm_audio_speech
  rest: POST /odbm/v1/audio/speech
  description: Compute speech biomarkers from an uploaded audio file.
  inputs:
  - file_url
  - platform
  - variables
caveats:
- >-
  Every analysis call is synchronous and long-running - the service processes the media inline via
  OpenFace in Docker and streams back a ZIP. An MCP tool binding would need an async job model that
  the API does not currently provide.
- >-
  Responses are ZIP archives of CSV, not JSON. A usable MCP tool would have to unpack and project
  them into structured output.
- >-
  The subject matter is video and audio of identifiable people, frequently clinical-trial
  participants. Any real MCP exposure of this surface would need an explicit consent and PHI-handling
  posture that does not exist today.