Skip to content
Docs/AI Control Centre/Create, version and retire AI agents

Create, version and retire AI agents

Agents with a role, instructions and a personality; every save is a new version, and any version can be restored.

Staff portalWho: Administrators and academic managersLast updated 3 September 2026
Flow

An agent's versions

Create AgentSave Changes → newversionVersion History:View / CompareRollback stages anold versionSave → new currentversion

Diagram: An agent's versions. Steps:

  • Create Agent
  • Save Changes → new version
  • Version History: View / Compare
  • Rollback stages an old version
  • Save → new current version

Connections:

  • Create Agent → Save Changes → new version
  • Save Changes → new version → Version History: View / Compare
  • Version History: View / Compare → Rollback stages an old version
  • Rollback stages an old version → Save → new current version
Staff actionPallara does this
History is never overwritten — a restore is itself a new version.
  1. Open AI → Agents and select Create Agent. Enter a unique name, choose the role, provide instructions and a personality, and set the primary and fallback models. Select Create Agent.
  2. Change the instructions and Save Changes. Agent updated — and a new immutable entry appears under Version History.
  3. Open Version History to View or Compare versions. Rollback stages a prior version in the editor; Save Changes then creates a new current version rather than overwriting history.
  4. Duplicate creates a Draft copy in its own editor. The row's archive action retires a single draft.
  5. Tick several cards for Activate Selected or Archive Selected; both confirm first and report a count. The status badge on each card changes immediately.
Agents: Aria, the learner AI tutor, and Navigator, the staff operations assistant.
FIG. 1Agents: Aria, the learner AI tutor, and Navigator, the staff operations assistant.

A new provider starts with two agents: Aria, the learner-facing AI tutor, and Navigator, the staff operations assistant behind the AI Assistant and the actions that arrive in Approvals. Each has a role, instructions, a personality and a model; you can create more.

Was this guide helpful?