Fleet Management

The Fleet view is the operational command center for all your agents. It gives you a live picture of what every agent is doing, how healthy it is, what it costs, and whether it needs attention.

The Workforce view

The primary fleet view shows all deployed agents in a card grid. Each card displays:

Indicator	What it shows
Status badge	Online / Offline / Degraded
Last active	When the agent last processed a message or cron job
Sessions	Number of open sessions
Daily tokens	Token consumption today vs. daily budget
7-day trend	Sparkline of activity over the past week

Click any agent card to open its detail view.

Agent status

Every agent has one of three statuses:

Online — The agent is running and responsive. Gateway is reachable, last heartbeat is recent.

Degraded — The agent is running but experiencing issues: elevated latency, repeated tool call failures, approaching budget limits, or session bloat causing slow responses.

Offline — The agent’s gateway is unreachable. This indicates a container crash, network issue, or the agent was stopped.

Agent detail view

Clicking an agent opens its full detail view with four tabs:

Overview tab

Current status and uptime
Model in use
Total lifetime tokens and estimated cost
SOUL file summary
Configured tools

Activity tab

Timeline of recent sessions and cron job executions
Token consumption per event
Tool calls made (type and count)
Delivery status for cron jobs

Sessions tab

List of all open sessions with message count, token size, and last activity
Archive — closes a session and removes it from active context (preserves history)
Prune — bulk archive of sessions older than a configurable threshold

Settings tab

Edit configuration (model, tools, limits)
Manage SOUL file
Configure cron jobs
Delete agent

Fleet-level metrics

The Fleet dashboard header shows aggregate metrics across all agents:

Total agents — deployed count
Active today — agents that processed at least one message or cron job today
Fleet daily tokens — combined consumption across all agents
Estimated monthly cost — based on current daily run rate
Alerts — agents flagged as Degraded or Offline

Managing sessions

Sessions are the most common operational lever you’ll use. Long sessions accumulate context, which slows response times and increases cost per turn.

Signs of session bloat:

An agent that was fast is now noticeably slower
Token consumption per cron job has crept up over weeks
A session has hundreds of messages and months of history

Best practice: Archive sessions that haven’t been used in more than 30 days. For cron job agents, their isolated sessions auto-close after each job — only review chat sessions manually.

To bulk-prune sessions older than 30 days:

Open the agent detail → Sessions tab
Click Prune Sessions
Set threshold (e.g., 30 days inactive)
Confirm — sessions are archived, not deleted

Health alerts

AgenFleet monitors agent health automatically and sends alerts when:

An agent goes Offline (heartbeat missed for >5 minutes)
An agent reaches 80% of its daily token budget (configurable threshold)
A cron job fails to deliver output for 2 consecutive runs
An agent’s response latency exceeds 30 seconds consistently

Alerts are delivered to the notification channel configured under Settings → Notifications.

Scaling the fleet

Adding more agents is straightforward — each agent is independently deployed and doesn’t affect others.

When to add an agent vs. give more work to an existing one:

Add a new agent when the task has a distinct identity, tool set, or audience (e.g., a compliance-focused agent vs. a research agent)
Expand an existing agent’s cron jobs when the task is similar in nature and the agent already has the right tools

There is no limit on fleet size in Operational and Enterprise plans.