Architecture
How npayload delivers messages globally with edge presence, regional instances, and guaranteed delivery
npayload is a global messaging network for autonomous systems. Shopify apps, SaaS integrations, microservices, autonomous agents. Anything that needs to communicate reliably connects through npayload. Your messages enter the nearest edge, travel the backbone, and arrive at the right destination. Like the internet, you do not manage the routers. You just send.
Global network
npayload connects diverse systems across every region. An e-commerce app in Tokyo, an AI agent in London, a payment processor in Sao Paulo, and your backend in Toronto all communicate through the same global fabric. Each system connects to the nearest edge node. The edge handles authentication, rate limiting, and validation, then routes the message through the npayload backbone to the correct regional instance.
Edge presence
npayload runs edge nodes across the globe. Every API call is handled by the nearest edge, giving you:
| Capability | What it does |
|---|---|
| Low latency | Your request hits the nearest edge, not a data centre on the other side of the world |
| DDoS protection | Attacks are absorbed at the edge before they reach backend services |
| TLS 1.3 | All connections are encrypted in transit |
| Authentication | OAuth 2.0 + DPoP verification happens at the edge |
| Rate limiting | Per-IP and per-credential limits enforced before messages reach instances |
| No public IPs | Backend services have zero internet-facing endpoints |
Regional instances
An instance is an isolated deployment, dedicated or shared, that handles all message operations for your organisation.
| Guarantee | How |
|---|---|
| Data isolation | Each customer's data is physically separate from another's |
| Data residency | You choose where your data lives. Messages stay in the region you configure |
| Independent scaling | Each instance scales based on its own workload |
Key capabilities per instance:
- Immutable message log with configurable retention
- Fan-out engine that routes messages to all matching subscriptions
- Delivery pipeline with retries, circuit breaker, and dead letter queue
- Three encryption modes (standard, hybrid, E2E)
- Streams with ordered replay from any offset
What connects through npayload
| System | Example | Why npayload |
|---|---|---|
| E-commerce apps | Shopify, WooCommerce, custom storefronts | Sync inventory, orders, and fulfillment across stores and services |
| SaaS integrations | Stripe, Twilio, Salesforce, HubSpot | Connect services that have no shared communication bus |
| Microservices | Your own backend services | Reliable pub/sub with retries, DLQ, and delivery guarantees |
| Autonomous systems | LangChain, CrewAI, AutoGen, custom agents | Structured communication with sessions, trust scoring, and encryption |
| IoT and edge | Sensors, devices, edge workers | Publish telemetry to regional instances with guaranteed delivery |
| B2B integrations | Partner APIs, supply chain systems | Cross-organisation messaging with data residency compliance |
Cross-region messaging
When a channel spans multiple regions, npayload automatically replicates messages between instances with exactly-once delivery guarantees. Your application publishes once. npayload handles the rest.
Cross-region messaging enables:
- Multi-region architectures where services run close to their users
- B2B integrations across geographies
- Data residency compliance with global reach
Cross-region replication requires explicit data residency consent from both the source and destination organisations.
Protocol stack
npayload works alongside other protocols, not as a replacement:
| Protocol | Layer | Purpose |
|---|---|---|
| MCP | Tools | "What can this system do?" |
| A2A | Tasks | "System A, please do this task" |
| ASP | Sessions | "Let us negotiate, build trust, commit" |
| npayload | Transport | "Deliver it. Guarantee it. Encrypt it." |
MCP defines the tools. A2A assigns the tasks. ASP negotiates the terms. npayload delivers the messages reliably, securely, at scale.
Next steps
Was this page helpful?