Crate aap_protocol 

AAP — Agent Accountability Protocol

The accountability layer MCP and A2A don’t have.

use aap_protocol::{KeyPair, Identity, Authorization, Level, Provenance, AuditChain, AuditResult};

let supervisor = KeyPair::generate();
let agent      = KeyPair::generate();

let identity = Identity::new(
    "aap://acme/worker/bot@1.0.0",
    vec!["write:files".into()],
    &agent, &supervisor,
    "did:key:z6MkSupervisor",
)?;

let auth = Authorization::new(
    &identity.id,
    Level::Supervised,
    vec!["write:files".into()],
    false, // not physical
    &supervisor,
    "did:key:z6MkSupervisor",
)?;

assert!(auth.is_valid());

Physical World Rule

let result = Authorization::new(
    "aap://factory/robot/arm@1.0.0",
    Level::Autonomous,   // Level 4
    vec!["move:arm".into()],
    true,                // physical = true
    &supervisor,
    "did:key:z6Mk",
);
assert!(matches!(result, Err(AAPError::PhysicalWorldViolation { .. })));

Structs

AuditChain

Tamper-evident append-only audit log.

AuditEntry

A single tamper-evident entry in the audit chain.

Authorization

Authorization token granting an agent permission to act.

Identity

An AI agent’s cryptographic identity. Signed by a human supervisor.

KeyPair

An Ed25519 keypair for signing AAP documents.

Provenance

Provenance records the immutable origin of an agent-produced artifact.

Enums

AAPError

All AAP errors implement this trait.

AuditResult

The result of an agent action.

Level

AAP authorization levels.

Functions

sha256_of

Compute "sha256:<hex>" of data.

verify_signature

Verify an Ed25519 AAP signature.

Type Aliases

Result