Receiving HL7 v2 over MLLP

MLLP over TCP is still how most HL7 v2 traffic moves inside a hospital. In IntuCloud it is the TCP source connector configured in MLLP mode.

Configuring the listener

The listener needs three things: a framing mode, a port, and what to do about acknowledgements.

• Framing — MLLP wraps each message in start and end blocks. Choose MLLP rather than raw TCP unless the sender is doing something unusual.
• Address — cloud listeners are reachable at a gateway address under *.gw.intucloud.ai. The gateway terminates TLS 1.2 or higher.
• Acknowledgement — the listener can generate an ACK automatically, or you can shape one in the postprocessor when the sender expects application-level detail.

ACK and NACK behaviour

The acknowledgement is a contract with the sending system, so it is worth being deliberate:

• AA (application accept) — the message was received and accepted.
• AE (application error) — the message was received but could not be processed. The sender will usually not retry.
• AR (application reject) — the message was rejected. Depending on the sender, this may or may not be retried.

When a channel is saturated, the listener returns a retryable NACK rather than accepting a message it cannot process. This is deliberate backpressure: the sending system is told to slow down, and nothing is silently dropped. HTTP listeners express the same behaviour as a 503 with a Retry-After header.

TLS and allowlists

Cloud MLLP listeners run over TLS. Plain MLLP on an untrusted network is not a supported cloud configuration.

Production listeners additionally require an inbound allowlist. You will need the sending system’s source addresses before a production listener will accept traffic — worth getting from the network team early, because it is a common cause of a stalled cutover.

For a stable address on the other side, dedicated ingress address sets are available as a $40/mo add-on and are included on Enterprise.

Testing before cutover

Send captured sample messages through the sandbox first. The trace shows the raw frame, the payload after each stage, and the acknowledgement that would go back to the sender. See Working with test data.