Monitoring
What It's For
Monitoring is where you check that your install is running and see what it has been doing. It has four tabs: Service Health (is each service up?), Events (the latest significant actions), Activity (every recent AI call and its cost), and PM Alerts (the time-tracking and budget alerts the PM Assistant sent). All four are read-only.
You find it in the Operator Console under System.
Service Health
Live check results for every monitored service on your install, problems first: down, then degraded, then misconfigured, then healthy. A summary line above the table counts each. A service that is switched off on your install, such as a retired job replaced by a newer one, has no check and is not listed or counted.
Each row shows the service's status, its name, the kind of check (Network check, Web check, Database check, or Activity check, which looks at how recently a scheduled job reported in), a detail line, whether it is specific to Agenteous or shared, and whether it is critical.
- Healthy: the check passed.
- Degraded: the service responded but not normally, or a scheduled job has not reported in as recently as it should.
- Down: the service did not respond or could not be found.
- Misconfigured: the check itself is not set up correctly, so the service is not being monitored. It does not raise an alert.
The data service that serves this page always shows as healthy, since it is the one answering. This tab answers "is it running?"; the Tech Stack screen in Agents and Tech Stack answers "what is installed?".
Events
The screen is titled Event Log. It shows the latest 200 significant actions across your install, newest first, looking back up to 30 days: settings saved, agent actions, rule decisions, and alerts. The tiles count Events, Event kinds, and Denied within what is shown.
Each row shows the time, the event (written as area and verb), the actor, the outcome (allowed, denied, or error), and the resource. Click a column header to sort. For a search by date range, actor, or outcome, use the Audit Log, which reads the same record.
Activity
Every AI call your install made recently, newest first: the latest 200, looking back up to 30 days. The tiles show Calls, Cost, Input tokens, and Output tokens (how much text the calls read and wrote). A By model breakdown shows each model's share of the cost.
The tiles and breakdown cover only the calls listed, not an all-time total. Each row shows the time, the model, the caller (which agent or job made the call), the input and output size, and the cost. Cost uses the provider's published price at the time; a model with no published price shows as $0.0000. For spend per day, use Spend and Time Saved.
PM Alerts
A record of the time-tracking and budget alerts the PM Assistant has posted to Slack, newest first. Each row is one alert: either a nudge to one person or a digest posted to your operations channel. This tab only records alerts; it does not send or clear them.
The tiles count Total Alerts, Alert Types, and Users (people who received a personal alert; channel digests are not counted). The table shows the time, alert type, user, task, status, channel, and a summary. If the PM Assistant has not sent anything, you see "No PM alerts recorded yet."
Who Uses It
- Operations administrators: check service health after an alert, and trace what happened and when.
- Agency owners: keep an eye on AI usage and on the time-tracking nudges your team receives.