Field selection

Record reads — the listings and the single-item GETs — accept ?fields= to return only the columns you will use:

curl -H "Authorization: Bearer $KOBANA_TOKEN" \
  -H 'User-Agent: My System (contact@example.com)' \
  -H "X-Workspace-Id: $WORKSPACE_ID" \
  'https://api.finance.kobana.com.br/v1/payables?status=pending&fields=description,amount,due_date'

{
  "data": [
    { "id": "8c8b4c4f-…", "description": "Rent", "amount": 1500.5, "due_date": "2026-11-10T00:00:00.000Z" }
  ],
  "meta": { "next_cursor": null }
}

Rules:

• Comma-separated names, in snake_case — the same ones the full response carries.
• Without the parameter (or with it empty), every field comes back.
• id always comes back, even when not requested.
• Top level only: asking for an object field (reconciliation, classification_centers) returns the whole object. There is no object.field syntax.
• A name the resource does not have → 400 BAD_REQUEST, with details.unknown (what was not recognised) and details.available (what exists). A wrong name is never silently ignored.
• A listing's meta is untouched — next_cursor is still there.
• Combines with filtering, sorting and pagination.

It does not apply to writes (which return the whole record), to totals (/aggregates, /cash-flow), to the id sweep (/ids) or to the attachment download — there the parameter is refused with 400.