Implementations Agentic AI

AgentDNA

AgentDNA gives autonomous agents a verifiable identity, explicit authority, and a record of what they actually did, the three things an agent needs before anyone can be accountable for it.

Visit AgentDNA (opens in a new tab) All implementations

AgentDNA
Domain
Agentic AI
Status
Ecosystem project
Built on
Identity + token chains

Overview

Identity, authorisation and provenance for AI agents.

AgentDNA states its scope in one line: identity, authorisation and provenance for AI agents. Those are three separate problems, and most agent stacks answer only the first (a key, a name, a session), while leaving the other two to logs written by the same system whose behaviour is in question.

Identity comes first because nothing else can be attributed without it. An agent registered through AgentDNA holds a decentralised identifier derived from the principal that created it, so its authority is always traceable to a person or an organisation that can answer for it.

Authorisation is issued as an explicit, scoped grant rather than an ambient credential. An agent cannot exceed a scope it was never given, and the boundary is stated in a record rather than implied by configuration.

Provenance closes the loop. Tool calls, data accesses and decisions anchor to independent Rubix token chains, so the account of what an agent did is not maintained by the agent, and withdrawing authority is itself a recorded event with verifiable ordering.

PRINCIPAL GRANT AGENT · SCOPED AGENT · SCOPED AGENT · REVOKED ANCHORED ACTION RECORD

What it uses the graph for

Four things it can prove.

01

Identity

Every agent holds a decentralised identifier derived from its principal, so an agent's authority is always traceable to someone who can be held responsible for it.

02

Authorisation

Permissions are issued as verifiable, scoped grants rather than ambient credentials. An agent cannot exceed a scope it was never given.

03

Provenance

Tool calls, data accesses and decisions anchor to independent token chains, so the record of what an agent did is not kept by the agent.

04

Governance and revocation

Withdrawing authority is an anchored event with verifiable ordering, which makes the line between authorised and unauthorised action demonstrable after the fact.

Live proof

AgentDNA, as it stands right now.

UnavailableAgents registered
UnavailableDelegations issued
UnavailableActions anchored
UnavailablePrincipals
UnavailableRevocations
UnavailableAvailability

No approved data contract for this implementation yet. Every tile shows the honest empty state rather than a placeholder number.