The agent header
| Control / Button | What it does |
|---|---|
| Back arrow | Returns to the previous screen. |
| Pause | Gracefully stops in-flight runs; queued runs wait. |
| Resume | Restarts queued and scheduled runs. |
| Restart | Terminates in-flight runs and starts fresh. |
| Status badge | Read-only indicator of the agent’s current state. |
The seven tabs
| Tab | What it’s for |
|---|---|
| Overview | Live snapshot of the agent’s health. |
| Activity | Timeline of everything the agent has done. |
| Memory | What the agent remembers, with cleanup tools. |
| Data Context | Data sources the agent can query and query history. |
| Outputs | Insights, alerts, and actions the agent has produced. |
| Evolutions | Version history and controls to improve the agent. |
| Configuration | The agent’s settings. |
Overview tab
| Control / Button | What it does |
|---|---|
| Total Runs | Read-only: how many times the agent has run, 7-day trend. |
| Success Rate | Read-only: share of runs that completed successfully. |
| Avg Execution | Read-only: typical time per run. |
| Confidence | Read-only: agent’s confidence in its results. |
| Agent details card | Identity and setup. |
| Current task card | What the agent is working on now. |
| Task queue card | Work lined up and waiting. |
| Resource usage card | Live compute and capacity. |
| Retry | Reloads the snapshot if it fails. |
Activity tab
| Control / Button | What it does |
|---|---|
| Kind filter chips | Event types: trigger, query, insight, alert, action, error, retry. |
| Since picker | Last hour, 24 hours, 7 days, 30 days, or All time. |
| Clear filters | Removes all active filters. |
| Row click | Opens a side panel with the full detail. |
Memory tab
| Control / Button | What it does |
|---|---|
| Memory stats tiles | Item and token counts per memory type. |
| Run prune | Drops low-relevance entries. |
| Run compact | De-duplicates and clusters similar entries. |
| Promote | Moves working memory into long-term storage. |
| Filter chips (source) | All, Conversation, Document, Tool, Web. |
| Type filter | Short-term, Long-term, Episodic, Semantic, Procedural, Working, Periodic. |
| Row selection | Select entries to act on in bulk. |
| Selection summary | Selected count and tokens. |
| Delete (bulk) | Permanently removes selected entries. No trash. |
| Export | Downloads selected memory as JSON or CSV. |
Data Context tab
| Control / Button | What it does |
|---|---|
| Generated context card | Agent’s summary of data it understands. |
| Connected data sources | One card per source with quality bar and reachable tables. |
| Source card click | Opens the data source detail page. |
| Browse data sources | Appears when no sources are connected. |
| Query History | Paged, time-stamped audit of queries. |
Outputs tab
| Control / Button | What it does |
|---|---|
| Insight filter bar | Filter by severity, kind, acknowledged-only. |
| Acknowledge (on an insight) | Marks an insight as reviewed. |
| Alerts & Notifications list | Delivery log of alerts. |
| Actions Executed list | Actions proposed or carried out. |
| Approve (on an action) | Approves a pending action; confirms first. |
| Reject (on an action) | Declines a pending action. |
Evolutions tab
| Control / Button | What it does |
|---|---|
| Auto-Evolve | Self-directed improvement run. |
| Targeted Tune | Focused improvement for a specific change. |
| Major Upgrade | Larger-scope improvement run. |
| Ready badge | Indicates the agent is available to evolve. |
| Evolution timeline | Chronological list of versions. |
| Version node click | Expands a past version. |
| Progress panel | Opens during an evolution run. |
Configuration tab
| Control / Button | What it does |
|---|---|
| Version history card | Digest of the three most recent versions. |
| Event triggers — Add trigger | Schedule, Data change, Webhook, or Manual. |
| Trigger row menu — Edit | Changes a trigger’s settings. |
| Trigger row menu — Test fire | Runs the trigger once. |
| Trigger row menu — Delete | Removes the trigger. |
| Copy webhook URL | Copies a webhook trigger’s address. |
| Security context — Edit | Opens the security editor. |
| Notification channels card | Manages alert destinations. |
| Safety policy card | Read-only mirror of the security setup. |
| Error log | Recent errors. |
Who can use it
- Org Owners and Admins have full reach: pause/resume/restart, edit security, approve actions, prune memory, launch evolutions.
- Department Admins and Managers operate agents in their team.
- Employees view Overview, Activity, Outputs, Data Context.
- Destructive actions always confirm; bulk memory deletion is permanent.