Glossary

A

A2A (Agent-to-Agent Protocol) - Google's open protocol for agent interoperability. Agents publish JSON Agent Cards at /.well-known/agent.json describing their capabilities. Attestix can generate and parse these cards.

Agent Card - A JSON document describing an AI agent's name, capabilities, URL, and version. Follows the A2A specification. Hosted at a well-known URL for discovery.

Attestation - A cryptographically signed statement about a fact. In Attestix, attestations include identity tokens, compliance declarations, and credential proofs.

B

Base L2 - A Layer 2 Ethereum rollup built by Coinbase. Attestix uses Base L2 for blockchain anchoring via EAS due to sub-$0.01 gas costs and fast finality.

Batch Anchoring - Combining multiple attestation hashes into a single Merkle tree and anchoring only the root hash on-chain. Reduces cost by up to 100x compared to individual anchoring.

C

Chain Hash - A SHA-256 hash linking each audit trail entry to the previous entry. Creates a tamper-evident log where altering any entry invalidates all subsequent hashes.

Conformity Assessment - An evaluation of whether an AI system meets EU AI Act requirements. Can be self-assessment (limited-risk) or third-party (high-risk via notified bodies).

D

DID (Decentralized Identifier) - A W3C standard for globally unique identifiers that do not require a central registry. Attestix supports did:key (ephemeral, no network needed) and did:web (domain-hosted).

DID Document - A JSON-LD document associated with a DID containing public keys, authentication methods, and service endpoints.

Declaration of Conformity - An Annex V document under the EU AI Act declaring that an AI system meets all applicable requirements. Attestix generates this as a W3C Verifiable Credential.

E

EAS (Ethereum Attestation Service) - An open protocol for creating on-chain attestations on Ethereum and L2 networks. Attestix uses EAS on Base L2 for tamper-proof anchoring.

Ed25519 - An elliptic curve digital signature algorithm (RFC 8032). Attestix uses Ed25519 for signing all identity tokens, credentials, and audit entries. Fast, compact, and quantum-aware.

Ed25519Signature2020 - A linked data proof type using Ed25519 signatures. Used in Attestix's W3C Verifiable Credentials.

EU AI Act - Regulation (EU) 2024/1689 establishing harmonized rules for AI systems in the European Union. Transparency enforcement begins August 2, 2026.

G

GDPR Article 17 - The "right to erasure" (right to be forgotten). Attestix implements this for agent identities, allowing cryptographic erasure while preserving audit trail integrity.

H

High-Risk AI System - An AI system classified under EU AI Act Annex III as requiring conformity assessment, registration, and ongoing monitoring. Examples: medical diagnosis, credit scoring, law enforcement.

J

JCS (JSON Canonicalization Scheme) - RFC 8785. A deterministic JSON serialization method ensuring the same data always produces the same byte sequence. Required for consistent signature verification.

JWT (JSON Web Token) - A compact, URL-safe token format. Attestix uses JWT for UCAN delegation tokens.

M

MCP (Model Context Protocol) - Anthropic's protocol for connecting AI models to external tools and data sources. Attestix is implemented as an MCP server exposing 47 tools.

Merkle Tree - A hash tree where each leaf is a hash of a data block and each non-leaf node is a hash of its children. Used in batch anchoring to efficiently anchor multiple attestations in a single transaction.

Multibase - A self-describing base encoding format. Attestix uses multibase-encoded public keys (e.g., z6Mk... for base58btc-encoded Ed25519 keys).

N

Notified Body - An organization accredited by an EU member state to conduct third-party conformity assessments for high-risk AI systems.

P

PBKDF2 - Password-Based Key Derivation Function 2. Used internally for deterministic key generation from passphrases.

Provenance - The documented history of an AI system's training data sources, model lineage, and decision-making process. Required by EU AI Act Articles 10, 11, and 12.

R

Reputation Score - A normalized 0.0-1.0 trust score calculated from an agent's verified interactions. Uses recency-weighted averaging (more recent interactions have higher weight).

Risk Category - EU AI Act classification: unacceptable (banned), high (Annex III, requires conformity assessment), limited (transparency obligations), or minimal (voluntary codes).

S

Schema UID - A unique identifier for an EAS attestation schema on-chain. Attestix uses a specific schema UID for its blockchain-anchored attestations.

U

UAIT (Unified Agent Identity Token) - Attestix's core identity primitive. A signed JSON document containing an agent's ID, capabilities, DID, metadata, and Ed25519 signature. Bridges MCP OAuth, A2A, and DID ecosystems.

UCAN (User Controlled Authorization Networks) - A decentralized authorization protocol using chained JWTs. Attestix implements UCAN v0.9.0 for capability delegation between agents with time-bounded, scope-limited permissions.

V

VC (Verifiable Credential) - A W3C standard for tamper-evident, cryptographically verifiable claims. Attestix issues VCs with Ed25519Signature2020 proofs for compliance declarations, identity attestations, and custom claims.

VP (Verifiable Presentation) - A W3C standard for bundling one or more VCs into a signed package for a specific verifier. Includes a challenge-response mechanism to prevent replay attacks.