Overview
The building blocks of the Agent Session Protocol
ASP is built from a small set of composable concepts. Each concept addresses a specific concern in agent-to-agent communication: how sessions are structured, how messages are formatted, how trust is measured, and how agreements are enforced.
Understanding these seven concepts gives you a complete mental model of how ASP works. They are listed in the order you will encounter them when building your first agent integration.
Quick reference
| Concept | What it covers | Key detail |
|---|---|---|
| Sessions | 8-phase lifecycle for every interaction | 9 states, configurable timeouts, multi-party support |
| Performatives | 13 message types with defined semantics | 4 categories: negotiation, information, action, lifecycle |
| Message format | JSON envelope, hash chain, signatures | SHA-256 integrity chain, Ed25519 signatures, DPoP binding |
| Trust scoring | 8-component reputation model | Logarithmic growth, exponential decay, breach drops |
| Commitments | Binding agreements with escrow | Irreversible after acceptance, automatic breach detection |
| Disputes | Resolution when commitments fail | Automated, peer-arbitrated, or human-escalated |
| Agent registry | Discovery and Agent Cards | Capability search, well-known endpoint, protocol interop |
Concepts
Sessions
The 8-phase lifecycle that structures every agent interaction, from discovery through learning
Performatives
The 13 message types agents use to propose, negotiate, commit, and close
Message format
The envelope structure, sequence numbering, and SHA-256 hash chain that ensure integrity
Trust scoring
The 8-component reputation model that quantifies agent reliability from observed behavior
Commitments
Binding agreements with escrow, penalties, and cryptographic proof of acceptance
Disputes
Resolution mechanisms for when agents disagree about whether commitments were fulfilled
Agent registry
Discovery, Agent Cards, and capability advertisement for finding the right counterparty
Was this page helpful?