generated: '2026-08-11'
method: derived
source: >-
  mcp/tadeus-api-mcp.yml (tool names published at https://tadeus.net/mcp) bound to
  openapi/tadeus-api-integration-openapi.json operationIds
note: >-
  The live MCP tools/list is auth-gated (401), so no tool inputSchema could be read.
  Every binding below is therefore matched by NAME AND SEMANTICS against the published
  REST operations, and confidence is set accordingly. Nothing here is asserted as a
  verified schema-level mapping.
surfaces:
  openapi:
    file: openapi/tadeus-api-integration-openapi.json
    spec: swagger 2.0
    base: https://app.tadeus.net/api/integration/v1
    operations: 47
    gated: true
    note: >-
      The spec document itself is public and anonymous; the operations require the
      X-API-KEY-ID / X-API-SECRET header pair.
  mcp:
    url: https://app.tadeus.net/mcp
    gated: true
    note: 401 on tools/list; tool set known only from the provider's product page.
  graphql:
    present: false
crosswalk:
  - tool: search_transcripts
    category: read
    rest:
      - transcripts_list
      - transcripts_read
    binding: semantic
    confidence: medium
    note: >-
      REST exposes transcript listing and retrieval but no documented full-text search
      parameter, so the MCP tool is at minimum a search layer over transcripts_list.
      The searching itself may be MCP-side.
  - tool: list_campaign_insights
    category: read
    rest:
      - insights_list
      - insights_read
    binding: semantic
    confidence: high
    note: >-
      insights_list accepts a campaign filter in the documented examples
      (params={"campaign_uuid": ...}), which is exactly this tool's shape.
  - tool: get_campaign_stats
    category: read
    rest:
      - campaigns_read
      - sessions_list
      - results_list
    binding: composite
    confidence: low
    note: >-
      No REST operation returns campaign statistics. "Completion by site" would have to be
      composed client-side from sessions_list (status filtering) and results_list. This is
      an MCP-side aggregation with no single REST equivalent.
mcp_only:
  - tool: get_campaign_stats
    reason: >-
      No REST operation returns aggregate campaign statistics; the MCP server computes it.
rest_only:
  - operations:
      - templates_list
      - templates_create
      - templates_read
      - templates_update
      - templates_partial_update
      - templates_delete
    reason: >-
      Interview template authoring has no named MCP tool. The MCP page describes an agent
      "defining the question", which implies a template-creating tool exists but is unnamed.
  - operations:
      - campaigns_list
      - campaigns_create
      - campaigns_read
      - campaigns_update
      - campaigns_partial_update
      - campaigns_delete
      - campaigns_invite
      - campaigns_bulk_invite
      - campaigns_generate_insights
    reason: >-
      Campaign lifecycle and invitation. The MCP page says agents can "commission
      interviews" and "monitor campaigns", so tools likely exist here but are not named.
  - operations:
      - sessions_list
      - sessions_create
      - sessions_read
      - sessions_update
      - sessions_partial_update
      - sessions_delete
    reason: Session lifecycle; no named MCP tool.
  - operations:
      - results_list
      - results_read
      - results_export_export_csv
      - results_export_export_json
    reason: >-
      Structured results and bulk export; no named MCP tool. Export-to-file is a poor fit
      for a tool response, which may be deliberate.
  - operations:
      - insights_create
      - insights_update
      - insights_partial_update
      - insights_delete
      - insights_html
      - insights_lexical
      - insights_markdown
      - insights_pdf
    reason: Insight authoring and the four export renderings; no named MCP tool.
  - operations:
      - attempts_list
      - attempts_read
    reason: Per-attempt records; no named MCP tool.
  - operations:
      - organisation_list
      - organisation_read
    reason: Account context; no named MCP tool.
  - operations:
      - billing_balance_list
      - billing_balance_read
      - billing_payments_list
      - billing_payments_read
      - billing_transactions_list
      - billing_transactions_read
    reason: >-
      Credit balance, payments and transaction ledger. Deliberately absent from the agent
      surface would be a defensible design choice, but it is not stated either way.
coverage:
  mcp_tools_published: 3
  mcp_tools_bound_to_rest: 2
  mcp_tools_mcp_only: 1
  rest_operations_total: 47
  rest_operations_with_a_named_tool: 8
  rest_operations_without_a_named_tool: 39
  confidence_note: >-
    The 39 uncovered REST operations are an upper bound on the divergence, not a measured
    one: the real tool list is gated and Tadeus names only 3 of its tools publicly.