generated: '2026-08-17'
method: derived
source: mcp/pyannoteai-mcp.yml + openapi/pyannoteai-api-openapi.yml
headline: >-
  pyannoteAI's MCP surface and its REST surface do not overlap at all. The MCP server is a Mintlify
  documentation-retrieval server; the REST API is the product. Zero of ten REST operations have an MCP
  tool, and zero of three MCP tools are backed by a REST operation. An agent connected to the MCP server
  can read about diarization but cannot perform it.
surfaces:
  openapi:
    file: openapi/pyannoteai-api-openapi.yml
    original: openapi/_original/pyannoteai-openapi.json
    source: https://docs.pyannote.ai/openapi.json
    server: https://api.pyannote.ai
    operations: 10
    gated: false
    note: Spec is publicly fetchable; calling the operations requires a bearer API key.
  asyncapi:
    file: asyncapi/pyannoteai-streaming-asyncapi.yml
    source: https://docs.pyannote.ai/asyncapi.yaml
    operations: 5
    note: WebSocket streaming gateway; not represented in MCP either.
  graphql:
    present: false
  mcp:
    url: https://docs.pyannote.ai/mcp
    tools: 3
    gated: false
    note: Anonymous tools/list succeeded on 2026-08-17; input schemas below are the REAL published ones.
crosswalk: []
mcp_only:
- tool: search_pyannote_ai
  reason: >-
    Documentation search over the published site. No REST operation exposes docs search — this is a
    Mintlify platform capability, not a pyannoteAI API capability.
- tool: query_docs_filesystem_pyannote_ai
  reason: >-
    Read-only virtual filesystem over documentation pages and specs. No REST equivalent.
- tool: submit_feedback
  reason: >-
    Writes documentation feedback to the docs team. No REST equivalent; not part of the product API.
rest_only:
- capability: Batch speaker diarization
  operations:
  - operationId: diarize
    method: POST
    path: /v1/diarize
- capability: Speaker identification with voiceprints
  operations:
  - operationId: voiceprint
    method: POST
    path: /v1/voiceprint
  - operationId: identify
    method: POST
    path: /v1/identify
- capability: Job lifecycle
  operations:
  - operationId: getJobById
    method: GET
    path: /v1/jobs/{jobId}
  - operationId: getJobsByTeamV2
    method: GET
    path: /v2/jobs
- capability: Temporary media storage
  operations:
  - operationId: getMediaUploadURL
    method: POST
    path: /v1/media/input
  - operationId: getMediaDownloadURL
    method: POST
    path: /v1/media/output
- capability: Real-time streaming diarization
  operations:
  - operationId: createStream
    method: POST
    path: /v1/live
  - operationId: getStream
    method: GET
    path: /v1/live/{id}
- capability: Connectivity check
  operations:
  - operationId: testKey
    method: GET
    path: /v1/test
coverage:
  mcp_tools_named: 3
  mcp_tools_bound_to_rest: 0
  mcp_only: 3
  rest_operations_total: 10
  rest_operations_with_a_tool: 0
  rest_operations_without_a_tool: 10
  binding_rate: 0.0
gap:
  summary: >-
    The highest-leverage agent-readiness gap for pyannoteAI. The provider has a clean 10-operation
    OpenAPI, a documented async job model and an HMAC-signed webhook callback — everything needed to
    forge a real operational MCP server — but the only MCP endpoint it publishes is the docs search that
    ships free with its Mintlify docs platform.
  recommended_tools_from_openapi:
  - tool: diarize_audio
    rest:
    - diarize
  - tool: create_voiceprint
    rest:
    - voiceprint
  - tool: identify_speakers
    rest:
    - identify
  - tool: get_job
    rest:
    - getJobById
  - tool: list_jobs
    rest:
    - getJobsByTeamV2
  - tool: upload_media
    rest:
    - getMediaUploadURL
  - tool: create_stream
    rest:
    - createStream
  note: >-
    Listed as a recommendation only. These tools DO NOT EXIST — they are not claimed anywhere in this
    repo as a pyannoteAI surface.