Activity & Memory

SuiteOp provides full visibility into what your agents are doing and what they have learned. This page covers how to monitor agent activity through sessions and action logs, and how the memory system works.

Sessions

Every time an agent processes a batch of events, it creates a session. Sessions are the primary audit trail for agent behavior.

Viewing Sessions

Navigate to the agent’s detail page and select the Sessions tab to see a chronological list of all sessions. Each session shows:

• Status — whether the session is running, completed, or failed
• Event count — how many events were in the batch
• Token usage — prompt tokens, completion tokens, and total
• Duration — how long the session took to process
• Reasoning — the agent’s summary of what it did

Session Details

Click on a session to see its full details:

• Events — the raw events that triggered this session, including their type, entity, property, and payload
• Actions — every tool call the agent made, with the tool name, arguments, result, permission mode, and approval status
• System prompt snapshot — the exact system prompt the agent received, useful for debugging
• Memories loaded — which memories were available to the agent during this session

Note

Sessions are retained for 7 days. Events are retained for 30 days. If you need longer retention, export the data before it is cleaned up.

Failed Sessions

If a session fails (due to an API error, timeout, or other issue), the events are released back to pending status and retried automatically. Events that fail 3 times are permanently skipped to prevent infinite retry loops.

Action Log

The action log is a cross-agent view of every tool call made by any agent in your organization. Access it from the Agents section to see all actions across all agents.

Filtering Actions

You can filter the action log by:

• Agent — narrow down to actions from a specific agent
• Approval status — filter by auto-executed, pending, approved, rejected, or expired
• Tool name — see all uses of a specific tool
• Date range — look at actions within a specific time period

Approval Statuses

Status	Meaning
Auto Executed	The tool was in Auto mode and executed immediately
Pending	The tool is in Ask mode and waiting for human approval
Approved	A human approved the action and it was executed
Rejected	A human rejected the proposed action
Expired	The action was not reviewed in time

Usage Summary

Each agent has a usage dashboard showing:

• Daily usage — tokens used today vs. the daily limit, plus session, event, and action counts
• Monthly usage — tokens used this month vs. the monthly limit, plus aggregate counts

This helps you monitor costs and adjust budget limits as needed.

Memory System

Agents have a shared memory system that allows them to learn and recall information over time. Memories persist across sessions and are shared between agents at the appropriate scope level.

Memory Scopes

Memories are organized in a hierarchy of five scopes:

Scope	Description	Example
Organization	Applies to the entire organization	”Our checkout time is 11 AM”
Property Group	Applies to a group of properties	”Downtown properties require noise monitoring”
Property	Applies to a specific property	”Unit 4B has a tricky thermostat that reads 3 degrees high”
Reservation	Applies to a specific reservation	”Guest requested late checkout”
Agent	Private scratchpad for a specific agent	Agent’s internal notes

When an agent processes events, it automatically receives memories from all relevant scopes — organization-level memories plus any that match the event’s property, property group, or reservation.

Memory Sources

Each memory has a source indicating how it was created:

Source	Description
Human set	Created or edited by a human user through the UI
Agent learned	Created by an agent during a session
Agent updated	An existing memory updated by an agent

Managing Memories

You can manage memories from the Memories section in the Agents area:

• Create — add a new memory at any scope level. Organization-level memories do not require a scope entity. All other scopes require you to specify the entity (property, property group, reservation, or agent).
• Edit — update a memory’s content or confidence level. Editing creates a new version — the previous version is preserved with a “superseded” status.
• Delete — soft-deletes the memory by marking it as superseded.
• Search — filter memories by scope and search by content.

Confidence Scores

Each memory has a confidence score from 0.0 to 1.0:

• 1.0 — fully confident (default for human-set memories)
• Lower values — the agent or user is less certain about this information

Confidence scores help agents weigh conflicting information. Human-set memories default to full confidence.

Memory Versioning

Memories use an append-only versioning system. When a memory is updated, a new version is created and the old version is marked as “superseded.” This preserves the full history of what was known and when, which is important for auditing agent decisions.

Note

Memories are presented to agents as “background knowledge” — clearly separated from the current events being processed. This prevents agents from confusing remembered facts with new events to act on.