Identity & change

Two rules, one consequence:

1. Every actor has an identity.
2. Changes are entities.

Put these together and provenance stops being a bolt-on audit log — it becomes native graph structure.

Every actor has an identity

The human owner, each AI agent, and the system itself are all entities on the graph. When the human edits a task, the change is attributed to them. When an agent creates a plan, it’s attributed to that agent. Every change has a who.

This is identification, not authentication — on a single-user local system, localhost binding is the access boundary. We’re not gatekeeping who can act; we’re recording who did.

• The human owner — one entity per AgentOS install. Your edits.
• AI agents — Claude, Cursor, your custom skills. Each a distinct entity.
• The system — automated processes like sync, indexing, migration. Attributed to a system actor, not hidden.

Changes are entities

When an entity is created, updated, or deleted, the operation itself becomes a change entity on the graph. A change has relationships to:

• the actor (who did it)
• the target (what changed)
• optionally the source (where data came from — a platform, a file, an API response)

This follows the pattern established by W3C PROV-O, ActivityStreams, and Git: make events first-class objects, not edges.

Provenance as traversal

Because changes are entities, provenance isn’t a static field — it’s the full chain of change entities. Walk backwards to reconstruct any previous state.

• “Who last edited this task?” — the newest change entity pointing at the task.
• “When did this email arrive?” — the creation change for the message entity.
• “Did the agent modify this, or did I?” — the actor on the most recent change.
• “Show me everything that happened yesterday” — query change entities by time.

No audit log table. No created_by / updated_at duplication per entity. One change shape, used universally.

Prior art

• W3C PROV-O — provenance ontology: agent, activity, entity. This is the pattern.
• ActivityStreams / ActivityPub — the fediverse’s answer to “what happened?” is an activity entity with an actor.
• Git — every commit is an entity with an author, a parent, and a diff. Same pattern, different domain.

See Research → Relationships and events for deeper background.

See also

• Design principles — “changes are entities” and “every actor has an identity” listed among the first-principles rules.
• Ontology overview — the change and actor shapes.