Pagination

Draft.

Keyset (cursor) pagination

List endpoints use cursor-based (keyset) pagination. There is no OFFSET or page number; this stays fast and consistent on millions of rows.

How it works:

1. You request the first page without a cursor.
2. The response returns an opaque cursor for the next page.
3. You request the next page by sending that cursor, and repeat until no cursor is returned.

Cursor rules:

• The cursor is opaque; do not try to decode or construct it.
• The cursor is versioned (it carries a v1. prefix). An undecodable or invalid cursor returns 400.
• Do not rely on a stored cursor remaining usable for a long time.

Parameter names

On the public API, parameter names will be in English. The names of the cursor and page-size parameters will be published on this page and in the API reference together with the first endpoints.

Tips

• Do not inflate the page size unnecessarily; for rate limits see Limits and versioning.
• If you want all data, follow the cursor; records added after your first request may appear on later pages depending on sort order.