generated: '2026-07-19'
method: searched
source: https://www.kita.ai/documentation
also_derived_from:
- openapi/kita-capture-openapi.yml
- openapi/kita-underwriter-openapi.yml
summary: 'Kita ships two independently-versioned REST APIs with different hosts, key prefixes
  and envelopes. Both are key-authenticated over TLS, both are asynchronous (submit then
  poll), and both use plain JSON rather than a hypermedia or problem-details format.'
authentication:
  style: bearer API key
  capture:
    header: 'Authorization: Bearer kita_prod_...'
  underwriter:
    headers:
    - 'Authorization: ApiKey kita_uw_...'
    - 'Authorization: Bearer kita_uw_...'
    key_scopes: [read, write]
  see: authentication/kita-authentication.yml
idempotency:
  supported: true
  api: Kita AI Underwriter API
  mechanism: client-supplied business key
  field: external_ref
  parameter_location: request body
  operation: POST /intake (openapi/kita-underwriter-openapi.yml#intakeApplication)
  behavior: 'Retrying an intake with the same external_ref returns the existing application
    with HTTP 200 and "idempotent": true rather than creating a duplicate, and skips
    re-uploading documents. A first-time intake returns 201.'
  header: none — Kita does not use an Idempotency-Key header
  retention: not documented
  scope: per organization
  capture_api: 'No idempotency contract is documented for the Kita Capture API; resubmitting
    the same file creates a new documentId.'
pagination:
  style: offset
  capture:
    endpoint: GET /api/v1/documents
    params: [limit, offset]
    defaults: {limit: 100, offset: 0}
    filters: [status, document_type]
    response_field: documents
  underwriter:
    endpoint: GET /applications
    params: [limit, offset]
    defaults: {limit: 50, offset: 0}
    limits: {min: 1, max: 200}
    filters: [status]
    response_envelope: '{ "data": [...], "pagination": { "total", "limit", "offset" } }'
  cursor_support: false
incremental_sync:
  supported: true
  api: Kita AI Underwriter API
  mechanism: 'GET /applications/{id}/conversation?after=<display_order> returns only messages
    newer than the supplied display_order.'
response_envelope:
  underwriter: 'Every response except binary exports is wrapped as { "data": ... }; list
    responses add a sibling "pagination" object.'
  capture: 'Result objects are returned unwrapped, sharing a common shape: status,
    document_type, document_id, filename, processing_time_seconds, uploaded_at, metadata,
    extracted_data and (when present) fraud_detection.'
error_envelope:
  underwriter: '{ "message": "..." } with a standard HTTP status.'
  capture: '{ "error": "Bad Request", "message": "..." }'
  rfc9457: false
  content_type: application/json
  see: errors/kita-problem-types.yml
async_processing:
  model: submit then poll
  capture:
    submit: 'POST /api/process-async returns { documentId, status: "pending" }'
    poll: 'GET /api/results/{documentId} until status is "completed"'
    statuses: [pending, processing, completed, failed]
    webhooks: 'Supported — register HMAC-signed endpoints, or pass a one-shot webhook_url on
      upload or batch creation.'
  underwriter:
    documents: 'Poll until document status moves from awaiting/processing to
      verified/low_confidence.'
    memo: 'Poll GET /applications/{id}/memo/status until synthesis_in_progress is false AND
      is_stale is false.'
    backoff: 'Kita recommends starting at 2–3 seconds with exponential backoff, not a tight
      loop.'
    webhooks: not supported — polling only
versioning:
  scheme: uri-path
  capture: 'Mixed — newer surfaces live under /api/v1/ (documents, batch, verify, webhooks,
    folders, schemas) while the original process and results endpoints sit at /api/.'
  underwriter: /api/v1
  breaking_changes: ship as a new version
  see: lifecycle/kita-lifecycle.yml
rate_limiting:
  capture:
    enforced: true
    scope: per organization
    signal: HTTP 429 with a Retry-After header
    published_numbers: false
  underwriter:
    enforced: not documented
    guidance: 'run document pushes at reasonable concurrency and poll on backoff (every few
      seconds, not a tight loop); coordinate heavy programmatic load with your Kita contact'
  see: rate-limits/kita-rate-limits.yml
request_tracing:
  request_id_header: not documented
field_expansion:
  supported: partial
  mechanism: 'GET /applications/{id}/documents?include=download_url adds one-hour signed
    download URLs. There is no general sparse-fieldset or expansion grammar.'
metadata:
  custom_metadata_field: false
  note: 'Underwriter applications carry external_ref for the caller''s own record ID and
    application_context for free text; Kita Capture offers folders and custom extraction
    schemas as the organizational primitives.'
content_types:
  request: [application/json, multipart/form-data]
  response: [application/json, text/plain, text/csv, application/pdf, application/vnd.openxmlformats-officedocument.spreadsheetml.sheet]
  base64_upload: 'Kita Capture accepts file_base64 with filename as an alternative to
    multipart; a data: URI prefix is accepted and stripped automatically.'
identifiers:
  underwriter_application: UUID or human app_id (e.g. APP-1234); both are accepted in the path
  capture_document: integer documentId
  capture_batch: string prefixed batch_ (e.g. batch_abc123)
data_residency:
  model: 'Region is determined by the organization''s configured residency, not by the API
    host. Request and response shapes are identical across regions.'
  regions: [ap-southeast-1, mx-central-1]
cost_reporting:
  supported: true
  api: Kita Capture API
  fields: [documents.total_cost_usd, documents.cost_report, batch_jobs.total_cost_usd]
  surfaced_on: GET /api/v1/documents/jobs/{documentId} and batch result endpoints
cross_links:
  errors: errors/kita-problem-types.yml
  lifecycle: lifecycle/kita-lifecycle.yml
  authentication: authentication/kita-authentication.yml
  rate_limits: rate-limits/kita-rate-limits.yml
  webhooks: asyncapi/kita-capture-webhooks.yml