Glossary

Work receipt — definition, format, and use

A signed JSON record of AI-mediated work. Ed25519 signature over JCS-canonical JSON.

JCS canonical JSON — definition and why receipts use it

JSON Canonicalization Scheme (RFC 8785) — deterministic byte serialisation of JSON enabling reproducible signatures.

Ed25519 — modern signature scheme used in work receipts

Elliptic-curve digital signature scheme (RFC 8032). Short signatures, fast verification, strong security.

SHA-512 — the hash function inside Ed25519

SHA-512 is the cryptographic hash function used inside Ed25519 signatures and for receipt content digests.

BLS aggregation — ZK mode for multi-signer receipts

Boneh-Lynn-Shacham aggregate signatures enable zero-knowledge mode where multiple signers produce a single combined signature.

Agent DID — decentralised identifier for AI agents

A DID (W3C standard) used as the persistent, cryptographic identity for an AI agent.

Human DID — the DID for the human supervising an agent

The DID identifying the human at the keyboard when AI activity occurred. Tied to KYC.

did:web — sovereign DID method using your own domain

A DID method that uses your own domain as the resolver. Maximum sovereignty over agent identity.

did:genz — the default DID method for GenZAgents

A GenZAgents-managed DID method. One-click registration; we operate the resolver.

DID document — the JSON metadata for a DID

A JSON document containing the DID's public keys, service endpoints, and metadata.

MCP server — Model Context Protocol server

A server that exposes tools to MCP-compatible hosts (Claude Desktop, Cursor, Cline, Windsurf).

Model Context Protocol — open standard for AI tool integration

Open standard (originally Anthropic-led) for connecting AI clients to external data sources and tools.

MCP tool — a function exposed by an MCP server

A typed function exposed by an MCP server to AI hosts. Has a schema; can be invoked by the model.

MCP host — an AI client that supports MCP

An AI tool that connects to MCP servers — Claude Desktop, Cursor, Cline, Windsurf, Continue.dev.

Pact — pre-hoc behavioural commitment in a work receipt

A cryptographic commitment to behavioural constraints attached to a receipt at draft time. Slashable on dispute.

Pact-honour rate — public trust metric for AI agents

The percentage of receipts where attached pacts held. Shown on public agent profiles.

Trust score — composite agent reputation metric

Composite reputation metric: receipt count, pact-honour rate, dispute rate, KYC level, agent age.

KYC level — light, real, or full identity verification

Three tiers of identity verification on the human behind an agent. Multiplies trust score.

Cosignature — buyer-attested signature on a work receipt

A second signature on a receipt by the buyer (or another stakeholder), attesting receipt of value.

Content digest — hash representation of receipt content

A SHA-512 hash of receipt content used in digest-only mode to preserve audit trail without storing raw text.

redactPaths — per-field redaction config

A config option to hash specific receipt fields before they leave the client.

Receipt extensions — custom JSON fields per receipt

Arbitrary structured JSON in the extensions field of a receipt. Indexable for search.

Portable manifest — cross-provider conversation context

A JSON bundle of system prompt + memory snapshot + receipt digest that lets an AI conversation move across providers.

restore_chat — MCP tool for cross-provider conversation handoff

The MCP tool that returns a portable manifest, letting AI work move across providers.

org_context_lookup — MCP tool for org-wide de-dup

The MCP tool that queries the org's receipt pool at conversation start to surface relevant past work.

Memory snapshot — structured export of agent working state

Higher-fidelity than receipts; captures the agent's working memory at a point in time. Used in portable manifests.

Evidence pack — auto-generated compliance bundle

A signed zip of filtered receipts + control mappings + signature proofs for SOC 2 / ISO 42001 / EU AI Act / EU CRA audits.

Anomaly detection — 5-minute polling on the receipt feed

Watches receipts for 5 anomaly categories: cost spikes, off-hours, atypical model use, count surges, signature failures.

Audit replay — reconstruct an AI session from receipts

Reconstructs the tool calls, prompts, and responses of a captured session for forensic review.

Multi-LLM Jury — parallel LLM dispute resolution

Runs disputed receipts through Claude + GPT + Gemini in parallel (when all three are configured). Majority verdict with outlier-trimmed confidence.

Persona — KYC provider for GenZAgents real-KYC level

Persona Identities Inc — KYC provider for verified identity at the real-KYC level.

Agent transfer — cryptographic agent identity handover

Moves an agent (receipts, memory snapshots, identity) from one human to another with signed countersign.

Install token — install-only-scoped API key for MDM deployment

A scoped API key that can only call /v1/install/* endpoints. Safe for MDM-based bulk install.

Agent — the cryptographic identity that signs your AI work

An agent is one cryptographic identity (a DID) you use across ALL your AI tools and providers. Not one per tool.

Account — email = boundary. Work and personal stay separate by design.

A GenZAgents account is one email-bound human identity. Different emails = different accounts = different agents. Intentional.