Agent Registration

Agent registration is a one-time setup process that provisions the infrastructure your agents need to store conversations, extract memories, and retrieve context. After registration, agents can immediately begin to ingest conversations and provide contextualized responses.

You can register agents programmatically using the Agentic Memory and Context APIs. To register an agent, you first create an agentic setup for your data space, then register the agent against that setup. For complete endpoint details, request schemas, and response formats, see Agent Configuration.

When you register an agent, Data 360 creates these configurations.

  • Agentic setup for the data space, if this is the first agent in the data space.
  • Ingestion API connectors and data lake objects (DLOs) to ingest raw conversations, memory, summaries, and execution traces.
  • Data model objects (DMOs) and mappings to store session-level data (raw conversations, short-term memory, summaries) and profile-level data (long-term memory).
  • Session data graph to capture session-level information including raw conversation data, short-term memory, and summaries.
  • Profile data graphs to capture long-term memory and references to sessions. Data 360 creates two out-of-the-box profile data graphs by default: one rooted in Individual and one rooted in User.

After registration, your agent moves through these status states.

StatusDescription
ProcessingRegistration in progress. The system is provisioning infrastructure. Agent not yet usable.
ActiveRegistration complete. All infrastructure is working normally. Agent ready for full operation.
ErrorAgent facing errors. Check underlying data graphs for error details.

You can delete an agent from the Registered Agents list or via the API.

Note: Deleting an agent removes agent-specific data including conversations, session history, and extracted memories associated with this agent. You can’t undo this action.

  • Data 360 derives agent status only from the out-of-the-box data graphs. If an additional real-time data graph is in error or still in creation, the agent remains active.
  • If an additional real-time data graph is unavailable during context retrieval, Data 360 returns data from out-of-the-box sources only and includes a warning in response metadata. The agent remains fully operational.