NewHow the Agent Session Protocol builds trust between autonomous systems.
Product

Your events reach every destination

Publish once. Every subscriber receives it with ordering, priority, and streaming built in.

You publish an event.

Ten services need it.

Some need it now.

Others can wait.

One subscriber fails.

The rest are stuck.

Retry logic in every service.

No ordering guarantees.

No dead letter queue.

You built plumbing instead of product.

Not anymore.

One channel type for every pattern

Classic pub/sub, ordered streams, or compacted channels. Choose the semantics that match your use case.

Pub/Sub

Publish an event. Every subscriber receives it independently. Slow consumers never block others.

Streams

Ordered consumption with resume from any offset. Your consumers never lose their place.

Compacted Channels

Only the latest value per key is retained. Ideal for configuration state and registries.

Publish once. Deliver everywhere.

One event reaches hundreds of subscribers. Slow consumers never block others. Failures route to the dead letter queue automatically.

Every message arrives in the right order.

Message groups provide strict FIFO ordering for events sharing a partition key. Events for the same order always arrive in publish sequence, regardless of throughput on other partitions. No complex rebalancing. No manual coordination. Ordering is guaranteed by the infrastructure.

Ordered Message Stream
1
order.created
order-123
seq:1
2
order.paid
order-123
seq:2
3
order.created
order-456
seq:3
4
order.shipped
order-123
seq:4
5
order.paid
order-456
seq:5
FIFO within partition key. Independent across keys.

Atomic publish across multiple channels

Publish to multiple channels in one operation. All channels receive the event or none do. Zero partial state in your system.

Step 1

Publish

Your service publishes to multiple channels in a single atomic operation.

Step 2

Validate

npayload verifies schema, permissions, and limits before persisting.

Step 3

Deliver

All channels receive the event or none do. Zero partial state.

Before and after npayload Messaging

Without npayload

  • Retry logic in every service
  • No ordering guarantees across subscribers
  • Lost events discovered when customers complain
  • No dead letter queue or a fragile homegrown one
  • Atomic multi channel publish is impossible
  • Months to build the distribution infrastructure

With npayload

  • Automatic retries with exponential backoff
  • Guaranteed ordering per partition key
  • Dead letter queue with one click replay
  • Fan out to hundreds of subscribers instantly
  • Native transactional multi channel publish
  • First event published in minutes

npayload vs. building it yourself

FeaturenpayloadBuild it yourself
Fan out to hundreds of subscribersMonths of work
Ordered streams with resumeComplex to build
Consumer group rebalancingMonths of work
Priority queues
Atomic multi channel publish
Dead letter queue with replayWeeks of work
Three privacy modes
Cross org delivery with consent
Zero infrastructure

Frequently asked questions

We already run Kafka. Why do we need npayload?+
Keep Kafka for internal services. npayload adds cross org delivery, encryption, consent, and audit trails. The Kafka adapter connects both.
Is there a message size limit?+
Up to 1 MB on all plans. Enterprise supports larger payloads. For very large data, publish a reference event with a signed URL.
How does ordering work when events fan out to many subscribers?+
Each subscriber receives events in publish order. Message groups provide FIFO ordering for events sharing a partition key.
What happens when a subscriber is down for hours?+
npayload retries with exponential backoff. If all retries are exhausted, events land in the dead letter queue for replay.
How does this handle spikes in traffic?+
Elastic scale. No capacity planning, no partition resizing. Traffic spikes are handled automatically.
Can we use npayload just for cross organization events and keep our internal broker?+
Yes. Many teams use npayload only for cross org events. Native adapters connect to your internal broker seamlessly.
What guarantees do you provide on delivery?+
At least once delivery. Every event is persisted before acknowledgment. Delivery tokens enable idempotent processing.
Is there vendor lock in?+
No. Standard HTTP and JSON. Plain REST API. Export your history and point publishers elsewhere at any time.