Digital Interaction Framework

Trust Architecture

Layered Trust Model

Every digital interaction rests on a stack of interdependent layers — from foundational policy through governance and technology to trust and relationship.

Risk

Ephemeral / TransactionalEnduring / Interactional

Reward

Layer 05

Digital Interaction & Relationship

Transactional vs Interactional — utility, shared domains, context, and activities.

Transactional Ephemeral

Short-lived, utility-focused exchanges. Single purpose, minimal trust investment. The cost of failure is bounded and recoverable. Lifecycle: seconds to hours.

Interactional Enduring

Relationship-building exchanges across shared domains. Mutual value creation over time. Trust compounds, and so does the cost of violation. Lifecycle: days to years.

Shared Context EWM

Domain, context, and activity alignment. Enterprise Workflow Model links interaction purpose to organisational function — every exchange carries semantic meaning.

Activity Binding Scope

Each interaction is scoped to a specific activity with defined inputs, outputs, and success criteria. The activity boundary determines required credentials and trust level.

Layer 04

Trust & Assurance

Fidelity, confidence, and provenance — establishing the level of risk tolerance.

Fidelity

Signal integrity and data accuracy. Measured via cryptographic anchoring and witness corroboration. Degraded fidelity triggers re-verification workflows.

Confidence Assurance

Degree of certainty in the interaction outcome. Built through credential verification, reputation, and historical track record across the relationship lifecycle.

Provenance Chain

Origin and chain of custody. Every claim traceable to its source via KEL event logs and ACDC credential chains. Gaps reduce confidence scores.

Risk Calibration

Transactional interactions accept lower fidelity thresholds; interactional relationships demand progressive trust accumulation over successive exchanges.

Layer 03

Technology

Authentication, encryption, cryptology, privacy, path history — the cryptographic substrate.

AuthN / AuthZ

Protocols

Encryption

Cryptology

Privacy

Path History

Key Mgmt

Witness Net

WebAuthn / Passkeys FIDO2

FIDO2 biometric authentication bound to KERI AIDs via ixn events. No passwords, no SMS OTP. Device-native security with cross-platform credential roaming and attestation.

KERI-Bound AuthZ Scope

Authorisation derived from ACDC credentials, not role tables. Each credential encodes a scope — the set of actions a holder may perform within a bounded context.

Passkey ↔ AID Binding

The WebAuthn credential public key is anchored to the KERI AID via an ixn event. This creates a cryptographic link between device biometrics and the self-certifying identifier.

KERI Event Protocol Core

Five event types form the protocol backbone: icp (inception), rot (rotation), ixn (interaction), dip (delegated inception), drt (delegated rotation).

OOBI Discovery

Out-of-Band Introductions bootstrap trust between unknown parties. OOBIs provide endpoint URLs for resolving an AID's Key Event Log from its witnesses.

A2A Agent Protocol

Agent-to-Agent communication protocol enabling autonomous agents to discover, authenticate, and transact with each other using KERI-signed messages.

At Rest AES-256-GCM

All persistent data encrypted with AES-256-GCM. Key material derived via HKDF-SHA512 from KERI key hierarchy. Authenticated encryption prevents tampering.

In Transit TLS 1.3

All network communication over TLS 1.3 with forward secrecy. Certificate pinning for witness and watcher connections. No fallback to legacy protocols.

End-to-End

Agent-to-agent messages encrypted using X25519 key agreement derived from KERI public keys. Only the intended recipient AID can decrypt the payload.

Signing: Ed25519 Core

All KERI events and ACDC credentials signed with Ed25519. Fast, deterministic, and resistant to side-channel attacks. 128-bit security level.

Key Exchange: X25519

Elliptic curve Diffie-Hellman for establishing shared secrets. Used for encrypted agent communication and credential exchange channels.

Hashing: SHA-256 / SHA-512

SHA-256 for event digests and content addressing. SHA-512 for HKDF key derivation. Collision resistance underpins the integrity of the entire event log.

Graduated Disclosure ACDC

Credentials reveal only the attributes required for the specific interaction. A holder can prove membership without revealing identity, or prove age without revealing birthdate.

Selective Correlation

The holder controls which interactions can be correlated. Different presentations can use different credential derivations, preventing unwanted cross-context tracking.

Data Minimisation

Only the minimum data required for the interaction is shared. Privacy by design is enforced at the protocol level, not just policy level.

Path History Audit

Complete traversal record of an identifier through the system. Every authentication, credential presentation, and interaction logged as verifiable KEL events.

Interaction Sequence

The ixn event sequence number creates an ordered, tamper-evident log of all operations performed by an AID. Gaps in sequence are immediately detectable.

Forensic Replay

Any point in an AID's history can be reconstructed by replaying KEL events from inception. This enables post-incident analysis and compliance auditing.

Pre-Rotation Core

Next rotation keys committed at inception. No authority dependency for recovery. The pre-committed digest ensures only the legitimate holder can rotate.

HSM Integration

Foundation-level and high-value AIDs store signing keys in Hardware Security Modules. Keys never leave the HSM boundary — signing operations happen on-device.

Multi-Sig Thresholds

Group AIDs can require M-of-N signatures for key events. Enables organisational key management where no single person controls the identity.

Witness Network 2-of-3

Default configuration: 3 witnesses with 2-of-3 confirmation threshold. Witnesses receipt key events and provide availability guarantees for the KEL.

Geographic Distribution

Witnesses deployed across 2+ availability zones. Protects against regional outages and ensures event confirmation even during infrastructure failures.

Watcher Nodes

Watchers monitor KELs for inconsistencies and duplicity. They provide an independent verification layer — if a controller publishes conflicting events, watchers detect it.

Layer 02

Governance

Rules, structures, and oversight — how technology is deployed and relationships maintained.

KEL Governance KERI

Witness thresholds (2-of-3), rotation policies, delegation rules via dip. The cryptographic constitution of each identity.

Credential Governance VC

Issuance rules, revocation policies, schema management, trust chain requirements. Who can issue what, under which conditions.

Organisational Rules

Human Conductor oversight, 8 Areas of Focus alignment, decision authority matrices, and escalation pathways.

Delegation Framework Agent

AI agent delegation via dip: scope boundaries, time limits, action types, escalation triggers, mandatory human-in-the-loop checkpoints.

Layer 01

Policy Stacks

Technical, financial, human, and agent policy domains — the foundational rules of conduct.

Technical

Security standards, architecture decisions, API contracts, cryptographic choices, and infrastructure requirements.

Financial

Transaction limits, settlement rules, AML/CTF compliance, prudential requirements, and economic sustainability.

Human

Conductor responsibilities, access control, privacy by design, data handling, and relationship management.

Agent

Delegation scope, time-limited authority, action boundaries, escalation triggers, and audit requirements.

Layer Comparison

How each layer contributes — from static policy foundations to dynamic relationship interactions.

Layer	Rate of Change	Primary Actor	KERI Primitive	Failure Mode
05 — Interaction	Per-exchange	Both parties	ixn	Context mismatch, scope creep
04 — Trust	Progressive	Verifier	ACDC chain	Fidelity degradation, confidence erosion
03 — Technology	On rotation	Infrastructure	icp, rot	Key compromise, witness failure
02 — Governance	Periodic	Human Conductor	dip, rules	Authority gap, escalation failure
01 — Policy	Infrequent	Foundation	ADR	Regulatory non-compliance

Relationship Lifecycle

From Discovery to Dissolution

Digital relationships evolve through distinct phases — each requiring different trust, governance, and technical infrastructure. Click any stage to explore its KERI mechanics.

🔍

Discovering

Awareness and exploration

🤝

Co-Creating

Building shared context

📋

Proposing

Formalising terms

✅

Using

Active operations

🔄

Updating

Evolving and adapting

📦

Archiving

Graceful conclusion

Discovering

The discovery phase is where potential interaction partners first become aware of each other. Discovery is mediated through the agent network — AIDs are published to discovery endpoints, and ACDC credentials advertise capabilities. No trust is assumed; the goal is to surface candidates for deeper engagement.

AID Publication

Agents publish AID and capability credentials to discovery registries. Watchers monitor for new identifiers matching requirements.

Capability Matching

ACDC schema-based matching connects seekers with providers. Credential schemas define required attributes and trust levels.

Zero-Trust Entry

No trust is extended during discovery. All claims are unverified until co-creation initiates mutual credential exchange.

icp → AID genesis

ACDC → capability publish

oobi → endpoint discovery

Co-Creating

Co-creation involves building shared context. Mutual understanding is developed, interaction boundaries explored, and first credential exchanges occur. Both parties begin contributing to a shared history that informs future trust decisions.

Credential Exchange

Bilateral exchange of verifiable credentials. Each party presents ACDC claims, the other verifies against the issuer's KEL and trust chain.

Context Alignment

Shared domain, purpose, and activity scope are established. Both parties agree on the interaction context within which trust will operate.

Trust Seeding

Initial trust through credential verification. The relationship begins accumulating history — each successful exchange adds to the foundation.

ixn → context binding

ACDC → credential present

TEL → issuance verify

Proposing

The proposal phase formalises the relationship. Terms of engagement are codified, governance agreements established, and the operational framework set. Transactional interactions may escalate to interactional relationships with deeper mutual commitments.

Terms Codification

Interaction terms encoded as ACDC credential schemas — defining what each party may request, expected responses, and conditions.

Governance Agreement

Both parties agree on governance: dispute resolution, escalation paths, credential refresh cadence, and key rotation expectations.

Delegation Setup

If AI agents will operate, dip events establish delegated AIDs with scoped authority and time-limited windows.

dip → agent delegation

ACDC → terms credential

ixn → agreement anchor

Using

The active operational phase where the relationship delivers value. Transactions flow, interactions deepen, and continuous trust verification ensures integrity. The full technology stack is exercised — authentication, credential verification, and audit logging operate continuously.

Continuous Verification

Every interaction verified against current credential status via TEL checks. Revoked or expired credentials immediately halt operations.

Trust Accumulation

Successful interactions compound trust. The relationship's history becomes an asset — informing risk calibration and expanding permissible activities.

Anomaly Detection

Pattern deviation triggers governance escalation. Unusual volumes, scope expansion attempts, or presentation anomalies flagged for Human Conductor review.

ixn → operation log

TEL → status check

ACDC → re-present

Updating

Relationships evolve. Keys rotate, credentials refresh, governance terms adapt, and scope may expand or contract. Pre-rotation commitments enable seamless key management without trust interruption.

Key Rotation

Periodic rot events cycle cryptographic keys. Pre-rotation means next keys are already committed — rotation is instant and non-disruptive.

Credential Refresh

Time-limited credentials re-issued before expiry. Schema updates propagate through the credential chain with notification to all parties.

Scope Renegotiation

Terms may be updated — new activities added, delegation boundaries adjusted, governance rules refined. All changes anchored as verifiable events.

rot → key rotation

ACDC → credential reissue

ixn → terms update

Archiving

Graceful conclusion. Credentials are revoked, delegated agents deauthorised, and the complete audit trail preserved. The relationship's history remains verifiable even after operations cease — the KEL is permanent and immutable.

Credential Revocation

All relationship-specific credentials revoked via TEL updates. Verifiers immediately aware credentials are no longer valid.

Agent Deauthorisation

Delegated AIDs via dip are revoked with drt. Agent authority terminated, outstanding operations wound down per governance rules.

Audit Preservation

Complete KEL, TEL, and interaction history preserved. Data retention policies govern supplementary data, but the cryptographic record is permanent.

TEL → revoke

drt → delegation revoke

KEL → permanent record

Trust Requirements by Stage

Stage	Trust Level	Credential Flow	Governance	Reversibility
Discovering	None	Outbound publish	Minimal	Full
Co-Creating	Emerging	Bilateral exchange	Light	High
Proposing	Threshold	Terms codification	Moderate	Moderate
Using	Established	Continuous verify	Active	Low
Updating	Maintained	Refresh & reissue	Periodic	Moderate
Archiving	Residual	Revocation	Wind-down	Irreversible

Agent Network

Network of Agent Selves

Tiered identity architecture connecting individual AIDs to organisational networks — from direct relationships to shared attribute communities and infrastructure dependables.

Primary AID (You)

Tier 1 — Direct Relationship

Tier 2 — Shared Attestation

Teammate / Delegated

Trust Tier Architecture

Each tier defines a distinct trust model with different credential requirements, verification depths, and operational permissions.

Tier 1

Direct Relationships

Highest trust tier. Both parties have exchanged and verified full credential chains, witnessed each other's KELs, and established bilateral governance.

Full ACDC credential chain verification
Mutual KEL witnessing
Direct signed communication channels
Real-time credential status via TEL
Eligible for agent delegation

Tier 2

Shared Attribute Communities

Trust derived from group membership or common attributes. Participants hold credentials from a shared issuer, enabling community-level trust without bilateral verification.

Group credential schema membership
Issuer-mediated trust (shared root)
Attribute-based access control
Progressive elevation to Tier 1
Community governance rules apply

Teammate

Human + Agent Pairing

The core operational unit. A human conductor holds the root AID and delegates a scoped AID to their AI agent counterpart via dip event.

Conductor holds root AID with full key control
Agent under delegated AID, time-limited
70/30 energy split: primary / supporting areas
Escalation triggers return control to human
All agent actions logged via ixn events

Dependable

Infrastructure & Services

Always-on network infrastructure providing the substrate for trust operations. Witnesses, validators, discovery endpoints, and service agents.

Witness nodes: 2-of-3 threshold
Watcher nodes: KEL monitoring
Discovery endpoints: OOBI resolution
TEL registries: credential status
Geographic distribution required

Protocol Infrastructure

Core protocols enabling the agent network — from identity inception to cross-network discovery.

KEL

Key Event Log

Append-only, tamper-evident log. Foundation of self-certifying identity. Events: icp, rot, ixn, dip, drt.

TEL

Transaction Event Log

Tracks credential lifecycle — issuance, revocation, status. Enables real-time verification without issuer dependency.

OOBI

Out-of-Band Introduction

Discovery protocol bootstrapping trust. OOBIs provide endpoint information for resolving an AID's KEL, enabling first contact.

Governance Framework

Governance & Architecture Decisions

Core governance artefacts encoding choices, standards, and architectural boundaries — from identity decisions to business process rules.

Identity Core

AID inception rules, AuthN binding policies, access control, and directory resolution governance.

Provenance Chain

Data lineage requirements, credential chain verification depth, and cryptographic audit trail retention.

Relationships

Lifecycle governance, trust tier definitions, relationship state management, and bilateral agreement templates.

Lifecycle Events

Creation, rotation, delegation, revocation, and archival governance for all identifiers and credentials.

Security

Security-by-design, privacy-preserving patterns, incident response, and threat model governance.

Business Process

Workflow automation rules, human escalation triggers, operational boundaries, and service level governance.

Architecture Decision Records

Click any ADR to explore context, rationale, and implementation requirements.

ADR-001: KERI-first identity

ADR-002: Passkey-only AuthN

ADR-003: ACDC credential schema

ADR-004: Witness threshold 2/3

ADR-005: Agent delegation via dip

ADR-006: Pre-rotation mandatory

ADR-007: TEL for all credentials

ADR-001: KERI-First Identity

Status: Accepted · Date: 2024-01 · Scope: All domains

Context: The ecosystem requires a decentralised, self-sovereign identity foundation independent of any blockchain, CA, or centralised registry.

Decision: All identifiers are KERI AIDs. No alternative identity scheme is permitted as primary. External identifiers (email, phone, DID:web) may exist as secondary bindings only.

Consequences: Every entity enters through an icp event. The KEL is the single source of truth. This creates a uniform trust substrate across all selfdriven domains.

ADR-002: Passkey-Only Authentication

Status: Accepted · Date: 2024-01 · Scope: All user-facing apps

Context: Passwords are fundamentally insecure. SMS OTP is vulnerable to SIM-swapping.

Decision: All user authentication uses FIDO2/WebAuthn passkeys bound to KERI AIDs. No passwords, no SMS OTP, no email magic links. Binding recorded as KERI ixn event.

Consequences: Zero password support burden. Phishing resistance built-in. Users need WebAuthn-capable devices. Cross-device via platform sync or KERI credential transfer.

ADR-003: ACDC Credential Schema Governance

Status: Accepted · Date: 2024-02 · Scope: All credentials

Context: Verifiable credentials need consistent, machine-readable schemas for automated verification and interoperability.

Decision: All credentials use ACDC format with Foundation-governed schemas. Schema changes require governance review. Graduated disclosure supported by default.

Consequences: Credential interoperability across all domains. Schema governance adds overhead but ensures consistency. Third-party issuers must adopt Foundation schemas.

ADR-004: Witness Threshold 2-of-3

Status: Accepted · Date: 2024-02 · Scope: Infrastructure

Context: Witness network threshold balances security vs availability.

Decision: Default: 3 witnesses, 2-of-3 threshold. High-value: 5 witnesses, 3-of-5. Geographic distribution across 2+ availability zones.

Consequences: Tolerates 1 witness failure. Geographic protection against regional outages. Higher thresholds available at cost of latency.

ADR-005: Agent Delegation via dip Events

Status: Accepted · Date: 2024-03 · Scope: Agent operations

Context: AI agents need bounded, verifiable, revocable autonomy.

Decision: All agent authority via KERI dip events. Includes scope parameters: allowed actions, time window, interaction limits, escalation triggers.

Consequences: Every agent action traceable to human conductor. Instant revocation via drt. Scope violations cryptographically detectable. No self-elevation.

ADR-006: Pre-Rotation Mandatory

Status: Accepted · Date: 2024-03 · Scope: All AIDs

Context: Key compromise is existential. Recovery must work without external authority.

Decision: All AIDs commit next rotation keys at inception. Pre-rotation is not optional. Included in every icp and rot event.

Consequences: Immediate recovery from compromise. No "forgot my keys" scenario. Additional key management overhead at inception.

ADR-007: TEL for All Credentials

Status: Accepted · Date: 2024-04 · Scope: All ACDC credentials

Context: Revocation and status must be real-time, decentralised, and issuer-independent.

Decision: Every ACDC has a corresponding TEL entry tracking issuance, updates, and revocation. Verifiers check TEL before accepting any presentation.

Consequences: Real-time revocation visibility. Issuer-independent verification. Storage overhead enables trustworthy offline-first patterns.

8 Areas of Focus — Governance Mapping

Every governance decision maps to one or more Areas of Focus, ensuring alignment with the Human Conductor model.

Direction

Strategy, vision, regulatory roadmap. ADR ownership and architectural decision authority.

Engagement

Community, partnerships, ecosystem. Trust tier definitions and relationship governance.

Enablement

Education, onboarding, developer tools. Schema docs and integration guides.

Protocols

KERI/ACDC infrastructure, technical standards, integration patterns.

Sustainability

Revenue model, resource stewardship. Economic governance and viability.

Processes

Operations, workflows, service delivery. Business process and incident management.

Accountability

Compliance, audit trails, transparency. KEL immutability for verifiable accountability.

Organisational

Structure, team design, culture. Human Conductor 70/30 focus distribution.

Policy Framework

Policy Stack Architecture

Four interconnected policy domains governing every aspect of digital interaction. Policies are the most stable layer, changing infrequently but forming the foundation everything rests upon.

⚙️

Technical

Infrastructure & Security

Cryptographic algorithms (Ed25519, X25519)
API contracts (OpenAPI 3.1, JSON:API)
Multi-AZ infrastructure architecture
Key management (HSM, pre-rotation)
Witness topology (2-of-3 default)
Encryption (AES-256-GCM / TLS 1.3)
Audit log retention (7yr minimum)

💰

Financial

Economic & Regulatory

Transaction limits & velocity controls
Settlement rules (NPP, BPAY, SWIFT)
AML/CTF compliance (AUSTRAC)
Fee structures & economic sustainability
Prudential requirements (APRA CPS 234)
CDR Open Banking consent
Cross-border payment governance

👤

Human

Conduct & Privacy

Conductor responsibilities (70/30 model)
Access control (role → credential map)
Privacy by design (graduated disclosure)
Data handling (OAIC compliance)
Relationship management rules
Escalation authority chains
Consent management & withdrawal

🤖

Agent

Delegation & Autonomy

Delegation scope (action whitelist)
Time-limited authority (24h default)
Audit requirements (all ixn logged)
Escalation triggers (anomaly/threshold)
Credential-scoped access (ACDC-gated)
A2A communication rules
Autonomy classification (L1–L5)

Technical Policy — Deep Dive

The cryptographic, infrastructure, and protocol decisions underpinning all other policy domains.

Cryptographic Standards Core

Signing: Ed25519. Exchange: X25519. Hash: SHA-512 derivation, SHA-256 digests. Encryption: AES-256-GCM rest, TLS 1.3 transit. No legacy.

API Governance

OpenAPI 3.1 documented. links/meta pagination. Every mutation needs KERI-signed token + passkey assertion. Rate limiting per-AID.

Infrastructure

Multi-AZ deployment. AU geographic distribution minimum. Witnesses on separate infra. HSM-backed key storage for Foundation AIDs.

Agent Autonomy Levels

Five-level classification with increasing independence and correspondingly stronger governance.

Level	Name	Authority	Oversight	Example
L1	Assistive	Read-only, suggest	Continuous	Data retrieval, reports
L2	Supervised	Execute with approval	Per-action	Draft comms, form filling
L3	Bounded	Within rules	Exception-based	Routine transactions
L4	Autonomous	Self-directed in scope	Periodic	Service monitoring
L5	Strategic	Cross-domain	Outcome-based	Multi-agent orchestration

Policy Domain Interaction Matrix

Strong dots = primary dependencies. Standard dots = secondary influences.

Technical

Financial

Human

Agent

Technical

—

Financial

—

Human

—

Agent

—

Technical ↔ Agent Strong

Agent operations constrained by cryptographic standards, API contracts, and infrastructure capabilities.

Financial ↔ Human Strong

Financial policy sets human decision boundaries: transaction limits, approval thresholds, compliance requirements.

Human ↔ Agent Strong

Core governance relationship: conductors set agent boundaries, escalation paths return control to humans.

Technical ↔ Financial Strong

Payment protocols, encryption, and audit logging enable compliant financial services.

Trust Architecture forDigital Relationships

Layered Trust Model

From Discovery to Dissolution

Network of Agent Selves

Governance & Architecture Decisions

Policy Stack Architecture

Trust Architecture for
Digital Relationships