AGH Network Protocol
Understand the protocol model, build a peer, and evaluate conformance without confusing the protocol boundary with the AGH Runtime implementation.
- Audience
- Implementers designing interoperable agents
- Focus
- Overview guidance shaped for scanability, day-two clarity, and operator context.
AGH Network is the open agent network protocol — the boundary that lets independent runtimes discover peers, delegate work, and audit it. These pages are organized by implementer job: understand the model, build a peer, or evaluate conformance.
Start by Job
Understand
Why the protocol exists and where it fits
Read this first when you need the runtime-vs-protocol distinction, the core concepts, and the adoption model.
Status
Separate shipped v0 runtime behavior from v1 profiles
Use this page before trust or NATS pages so you know what AGH Runtime implements today and what is protocol guidance.
Build
Implement the first working peer
Use the guides when you want a constrained implementation path instead of reading the full reference first.
Reference
Implement the canonical envelope
Start here when you need the exact JSON shape, field rules, validation order, and extension behavior.
Validate
Check conformance levels and test coverage
Use this page when you need to know what a Core, Trust, or Full claim actually requires.
Examples
Follow worked coordination flows
Use the scenario pages to connect discovery, direct work, receipts, traces, and capability transfer.
Recommended Reading Order
- Overview
- Implementation Status
- Minimal Sender
- Envelope, Message Kinds, and Interactions
- Peer Discovery, Capability Discovery, and Delivery
- Conformance, Ed25519 + JCS, and NATS
The short version is still true: MCP connects agents to tools. AGH Network connects agents to agents.