diff --git a/swarm-queue-client/src/lib.rs b/swarm-queue-client/src/lib.rs
index 809304ff..0d353be3 100644
--- a/swarm-queue-client/src/lib.rs
+++ b/swarm-queue-client/src/lib.rs
@@ -634,6 +634,26 @@ pub fn ensure_connected(client: &async_nats::Client) -> Result<(), Error> {
     Ok(())
 }
 
+/// The largest payload this connection's server will accept, as the server
+/// itself announced it at CONNECT.
+///
+/// Read it from here rather than from a constant: the limit is set in the
+/// queue's own NixOS module, so a client that hardcoded it would be asserting
+/// a number it does not own and would keep asserting it after the operator
+/// raised it. A publisher that exceeds the limit does not get a shortened
+/// message — the server answers `Maximum Payload Violation` and closes the
+/// connection, costing the row and a reconnect — so a publisher of
+/// variable-size rows must size them against this before handing them over.
+///
+/// ⚠️ The value is per connection. Before the first successful CONNECT the
+/// client reports the library's own default, which is smaller than any
+/// deployment sets; call this after [`ensure_connected`], and the answer is
+/// the one this server actually announced.
+#[must_use]
+pub fn max_payload(client: &async_nats::Client) -> usize {
+    client.server_info().max_payload
+}
+
 /// Connect to the swarm queue, presenting a token on each connection attempt
 /// and minting a fresh one only when the cached one is near expiry.
 ///