generated: '2026-07-25'
method: derived
source: mcp/ki-insurance-mcp.yml + openapi/ki-insurance-broker-platform-openapi.yml
status: candidate
summary: >-
  Ki has no published MCP server and no GraphQL surface, so this crosswalk binds
  the CANDIDATE tool set in mcp/ki-insurance-mcp.yml to the observed REST
  operations it would be built on. Because every candidate tool is derived from a
  REST operation, the binding is 1:1 and confidence is high by construction —
  there are no mcp_only rows. The value of the file is the rest_only section: it
  records how much of Ki's underwriting surface would remain unreachable to an
  agent, and why.
surfaces:
  openapi:
    file: openapi/ki-insurance-broker-platform-openapi.yml
    gated: true
    note: >-
      Derived from the published client bundle, not from Ki. Calling any
      operation requires partner Auth0 credentials.
  graphql:
    endpoint: null
    note: >-
      No GraphQL surface. /graphql on app.ki-insurance.com returns HTTP 200 with
      the SPA's index.html — a catch-all false positive, explicitly rejected.
  mcp:
    url: null
    gated: true
    note: No hosted MCP server exists; tools/list could not be called.
crosswalk:
- {tool: search_indications, category: discovery, rest: [getQuoteIndicationSuggestions], binding: rest, confidence: high}
- {tool: list_pipeline, category: discovery, rest: [getPipelineList], binding: rest, confidence: high}
- {tool: list_leads, category: discovery, rest: [getLeadsList], binding: rest, confidence: high}
- {tool: get_lead, category: discovery, rest: [getLeadsLead], binding: rest, confidence: high}
- {tool: get_quote, category: quote, rest: [getQuoteById], binding: rest, confidence: high}
- {tool: get_quote_detail, category: quote, rest: [getQuoteDetail], binding: rest, confidence: high}
- {tool: get_quote_document, category: quote, rest: [getQuotePdf], binding: rest, confidence: high,
   note: Returns a Blob rather than JSON; the client branches on Content-Type.}
- {tool: list_classes_of_business, category: reference, rest: [getClassOfBusiness], binding: rest, confidence: high}
- {tool: list_nominated_leads, category: reference, rest: [getSyndicateNominatedSyndicateByClassOfBusiness], binding: rest, confidence: high}
- {tool: list_risk_codes, category: reference, rest: [getRiskCodes], binding: rest, confidence: high}
- {tool: search_broking_houses, category: reference, rest: [getBrokingHouseSuggest], binding: rest, confidence: high}
- {tool: create_quote, category: quote, rest: [createQuote], binding: rest, confidence: high}
- {tool: create_quote_from_indication, category: quote, rest: [createQuoteFromIndication], binding: rest, confidence: high}
- {tool: update_quote, category: quote, rest: [updateQuoteById], binding: rest, confidence: high}
- {tool: create_quote_version, category: quote, rest: [createQuoteVersion], binding: rest, confidence: high}
- {tool: upload_schedule_of_values, category: exposure, rest: [createQuoteSov, updateQuoteSov], binding: rest, confidence: high,
   note: Multipart upload; POST creates and PUT replaces the same resource.}
- {tool: geocode_schedule_of_values, category: exposure, rest: [createQuoteProcessSov], binding: rest, confidence: high,
   note: Asynchronous job; no completion webhook exists, the caller polls.}
- {tool: start_pricing, category: pricing, rest: [updateQuoteStartPricing], binding: rest, confidence: high,
   note: "Asynchronous. Ki's own client polls quote detail every 5s for up to 90s (GET_PRICING_INTERVAL/GET_PRICING_TIMEOUT)."}
- {tool: discard_quote, category: quote, rest: [updateQuoteDiscard], binding: rest, confidence: high}
- {tool: extract_slip, category: ingestion, rest: [createSlipExtractionExecuteSlipExtractionJob], binding: rest, confidence: high}
- {tool: get_slip_extraction_status, category: ingestion, rest: [getSlipExtractionJobStatus], binding: rest, confidence: high}
- {tool: quote_from_slip, category: ingestion, rest: [createSlipExtractionQuoteFromJobResult], binding: rest, confidence: high}
- {tool: issue_firm_order_note, category: bind, rest: [createQuoteFirmOrderNoteV2], binding: rest, confidence: high,
   note: Human-in-the-loop required — this commits syndicate capacity.}
- {tool: export_pipeline_csv, category: reporting, rest: [createPipelineCsv], binding: rest, confidence: high,
   note: "POST with a JSON body and Accept: application/csv."}
mcp_only: []
rest_only:
- capability: Master data / reference lookups
  count: 38
  examples: [getMasterdataPropertyPerils, getMasterdataCyberKiLimitbasis, getMasterdataHqCountries,
    getMasterdataTerritories, getMasterdataCompliance]
  reason: >-
    Form-driving enumerations. Better exposed to an agent as MCP resources than
    as tools; excluded from the candidate tool set deliberately.
- capability: Platform administration
  count: 6
  examples: [getAdminSettings, updateAdminSettings, getAdminMaintenance,
    updateAdminMaintenance, createAdminUploadMaxLineSize, createAdminValidateMaxLineSize]
  reason: Operator surface, including max line size — the control that bounds how much
    capacity Ki will write. Never appropriate for an agent.
- capability: User and broking-house administration
  count: 17
  examples: [getUserAll, createUserAll, updateUserWithId, updateUserClassOfBusiness,
    deleteUser, createUserBulkUpload, createBrokingHouseBranches, updateBrokingHouseBranch,
    deleteBrokingHouseBranch, getBrokingHouseLloydsBrokerId]
  reason: Identity and entitlement management, not underwriting.
- capability: Dashboard and facilities
  count: 7
  examples: [getDashboardIndications, getDashboardRenewals, getFacilitiesLineslipMappings,
    getExclusionClausesList, getFacilityAttestations, getSyndicateTrackerSyndicates,
    getConfigLineslips]
  reason: Useful but secondary; candidates for a second tool wave.
- capability: Telemetry
  count: 2
  examples: [createFrontendMetrics, createUserJourneyEvents]
  reason: Client instrumentation, not a capability.
coverage:
  tools_named: 24
  tools_bound: 24
  mcp_only: 0
  rest_operations_total: 109
  rest_operations_with_a_tool: 25
  rest_operations_without_a_tool: 84