generated: '2026-07-27'
method: derived
source: mcp/hydro-quebec-mcp.yml + openapi/hydro-quebec-open-data-explore-api-v2-1-openapi.json
note: >-
  Binds each Opendatasoft MCP tool exposed by the live Pipeworx gateway to the OpenAPI operation
  that backs it in the Hydro-Québec Explore API v2.1 description. The MCP tool schemas here were
  read from a real tools/list response, so every binding is name-and-parameter verifiable rather
  than guessed. Confidence is high where the tool's arguments map one-for-one onto the operation's
  declared parameters, medium where the tool uses legacy Search API v1 argument names (rows/start)
  that the v2.1 operation expresses as limit/offset.
surfaces:
  openapi:
    file: openapi/hydro-quebec-open-data-explore-api-v2-1-openapi.json
    base_url: https://donnees.hydroquebec.com/api/explore/v2.1
    operations: 16
    gated: false
  graphql:
    endpoint: null
    note: Hydro-Québec publishes no GraphQL surface; the Opendatasoft platform does not expose one.
  mcp:
    url: https://gateway.pipeworx.io/opendatasoft/mcp
    official: false
    gated: false
    note: tools/list returned HTTP 200 anonymously on 2026-07-27
crosswalk:
- tool: datasets
  category: catalog
  rest: [getDatasets]
  binding: rest
  confidence: medium
  note: >-
    Tool arguments q/rows/start/sort/facet are the Search API v1 spellings; getDatasets expresses
    the same contract as where/limit/offset/order_by/refine in v2.1 ODSQL.
- tool: dataset
  category: catalog
  rest: [getDataset]
  binding: rest
  confidence: high
  note: dataset_id maps directly onto the {dataset_id} path parameter.
- tool: records
  category: data
  rest: [getRecords]
  binding: rest
  confidence: high
  note: >-
    where/select/group_by/order_by/limit/offset are the exact ODSQL parameter names getRecords
    declares in components.parameters.
- tool: facets
  category: data
  rest: [getRecordsFacets]
  binding: rest
  confidence: high
  note: dataset_id + facet map onto {dataset_id} and the facet parameter.
- tool: instance_info
  category: catalog
  rest: [getDatasets, getDatasetsFacets]
  binding: rest
  confidence: low
  note: >-
    Composite — catalog-level counts/themes/languages are assembled from the catalog listing and
    catalog facets rather than from a single dedicated operation.
mcp_only: []
rest_only:
- capability: catalog export
  operations: [listExportFormats, exportDatasets, exportCatalogCSV, exportCatalogDCAT]
  reason: >-
    Bulk/format exports of the dataset catalog (CSV, DCAT RDF/XML) have no MCP tool; agents must
    call the REST endpoints directly.
- capability: dataset export
  operations: [listDatasetExportFormats, exportRecords, exportRecordsCSV, exportRecordsParquet, exportRecordsGPX]
  reason: >-
    Unlimited bulk record exports (CSV, Parquet, GPX) have no MCP tool. This is the significant
    divergence — the MCP surface is limited to paged record queries, so large historical pulls
    (e.g. 2.3M weather rows) must go through the REST exports endpoints.
- capability: single record read
  operations: [getRecord]
  reason: No MCP tool reads one record by record_id.
- capability: dataset attachments
  operations: [getDatasetAttachments]
  reason: No MCP tool lists dataset attachments.
coverage:
  tools_named: 5
  tools_bound: 5
  mcp_only: 0
  rest_operations_total: 16
  rest_operations_with_tool: 5
  rest_operations_without_tool: 11