List outbound sent emails

Pagination cursor from a previous response's meta.cursor field. Format: {ISO-datetime}|{id}

Number of results per page

Filter to rows in this status. Useful for polling queued rows that haven't transitioned, auditing gate-denied attempts, or listing only successful deliveries.

Filter to the row matching a specific server-issued request_id. The /send-mail response surfaces request_id on every send; this lookup lets the caller find the historical row for a given live call without remembering its id.

Filter to rows with the given client_idempotency_key. Multiple rows can share a key (a retry that hit the idempotent-replay path returns the same row, but a retry with a DIFFERENT canonical payload under the same key is rejected by /send-mail before the row is written, so duplicates are bounded).

Inclusive lower bound on created_at. An ISO 8601 timestamp with Z or a numeric UTC offset (2026-10-02T00:00:00Z, 2026-10-02T00:00:00-04:00). An offset is read as the instant it names. A time with no zone, or a bare date, is rejected.

Inclusive upper bound on created_at. An ISO 8601 timestamp with Z or a numeric UTC offset (2026-10-02T00:00:00Z, 2026-10-02T00:00:00-04:00). An offset is read as the instant it names. A time with no zone, or a bare date, is rejected.

Literal case-insensitive substring filter over the subject, the retained plain-text body, the sender, and the To, CC and BCC recipients of each send. Matched verbatim: %, _ and \ are ordinary characters and do not act as wildcards, so a search for 10% finds the literal string 10% rather than everything beginning 10.

Leading and trailing whitespace is trimmed, and the result must be 3 to 500 UTF-16 code units. The three-character minimum is the search index's: a shorter pattern cannot be answered from the index and would scan the whole mailbox, so it is rejected rather than served as a request that times out.

Those bounds are deliberately NOT declared as minLength/maxLength. Both count Unicode code points on the value as sent, while the server trims first and counts UTF-16 code units, and the two disagree in both directions: two astral characters are 2 code points but 4 code units, so a declared minimum would reject a query the server accepts, and "<499 chars> " is 500 code points before trimming and 499 after, so a declared maximum would reject another. Declaring a bound the server does not enforce is worse than declaring none, because the client refuses locally and the caller never learns why. The server validates; a violation is a 400.

Recipients are matched one address at a time rather than against the serialized recipient list, so a query containing a quote, comma or bracket matches only where that character occurs inside an address. Bodies are searched only where the plain-text body was retained; HTML-only bodies are not searched.

Not served by every deployment. Where sent-mail search is not enabled the endpoint answers 503 with code sent_mail_search_unavailable, which callers should treat as a capability answer and fall back on rather than as an outage.

Exact case-insensitive sender mailbox filter. Returns only sends whose from address equals this value; it is not a substring match, so a partial mailbox matches nothing. Use q to search sender text loosely. Combines with every other filter, and unlike q it is served by every deployment.