AI as First-Class KERI Actor — Specification

Type: Design specification

Priority: 🟢 Phase 1 (Months 3–6)

---

Purpose

Define how AI agents operate as delegated KERI participants.

Key Positioning

“AI that can legally and cryptographically act, not just suggest.”

Core Concepts

AI Agents with Delegated AIDs

• Explicit authority scope
• Time-bounded delegation
• Revocable at any time
• Auditable action history

Non-Repudiable AI Actions

• Every action signed by agent’s AID
• Delegation chain verifiable
• Principal accountable
• Agent actions attributable

AI Capabilities

Execute Contracts

• Sign agreements within scope
• Honor pre-defined terms
• Trigger conditional actions

Negotiate Offers

• Propose terms within parameters
• Accept counter-offers if authorized
• Escalate when outside scope

Perform Disclosures

• Present credentials as authorized
• Selective disclosure by policy
• Audit trail of what was shared

Integrate Services via ACDC Schemas

• Read schema definitions
• Compose appropriate disclosures
• Establish OOBI trust
• Execute without pre-integration

Integration-on-the-fly Pattern

1. Read schema — Understand what’s being asked
2. Compose disclosure — Assemble required credentials
3. Establish OOBI trust — Bootstrap connection
4. Execute — Complete the interaction

No pre-built integrations required. AI reads the schema and figures out how to comply.

Delegation Patterns

Scope Limitations

• Action types allowed
• Value limits
• Time boundaries
• Counterparty restrictions

Revocation

• Immediate effect
• No delayed propagation
• Cryptographically enforced

Audit

• Complete action history
• Delegation chain visible
• Principal always identifiable

---

TODO: Write detailed specification