Unify Logo Footer.svg
Unify Agentic AI
Logo
Spending and Alerts

Spending and Alerts

Logo

3 mins READ

Spending and Alerts track an agent's cost, let you cap its spending, and can alert you — or shut it down — before it runs over budget.

Overview

The Spending tab in an agent's Observability area lets you "Monitor usage and manage spending limits." It shows how much the agent has cost this month, lets you set a monthly cap on the agent and a per-user cap, and lets you configure alerts that fire when cost crosses thresholds. The idea up front: set monthly caps, track spend in real time, get alerted near the limits, and optionally pause the agent automatically when a limit is reached.

Tracking Spend

The Spending tab shows the current-month spend as a dollar amount. Once you set a limit, it also shows a usage percentage and a progress bar with markers for each alert threshold, so you can see how close you are to the cap at a glance. If usage passes the limit, the indicator turns red and reads {usage}% Exceeded.

Setting a Spending Limit

You can set two kinds of monthly cap:

Limit type

Description

Action

Agent limit

Set a monthly spending limit for the agent to control costs — applies to total agent spending across all users.

Add Limit / Edit Limit

Per-user limit

Set a default monthly spending limit for each individual user to distribute costs fairly.

Add User Limit / Edit User Limit

Both open a small dialog with a Monthly Spending Limit field; leave it empty to remove the limit. Limits save as soon as you set them.

Configuring Spending Alerts

In the Configure alerts card, use Add New Alert — "Set threshold values and actions upon breaching them for agent spending." In the alert form, set:

  • Alert when agent costs reaches — the cost that triggers the alert.

  • Alert Interval — how often the alert is evaluated (a number plus Minutes, Hours, or Days).

  • Define Thresholds — per-status levels for Critical, Warning, and Resolved, each as a percentage of budget or a spending amount.

  • Define Actions — what happens at each status.

Each alert can be toggled on or off and edited or deleted from the card.

Alert Actions

When you add an action to an alert, you pick its Status (Critical, Warning, or Resolved) and an action type:

Action type

Description

Email

Send to one or more email addresses.

Notification

Post a platform notification to chosen Users or Teams.

Custom

Run a custom action (an app operation through a connection).

Undeploy Agent

Automatically take the agent down when the threshold is breached.

Undeploy Agent is the hard stop — pair it with the Critical status to pause an agent that has blown its budget, and use Email or Notification at Warning to give yourself a heads-up first.

What Undeploy Agent Actually Does

When the threshold breaches, the platform finds every live deployment of the agent and switches it to undeployed — the agent stops serving on all its channels at once. Nothing is deleted: the agent, its configuration, and its deployment records remain, so once you've addressed the spend you can redeploy from the Deployments screen. The other action types run in the same pass — platform notifications go to the chosen users and teams (membership is resolved at fire time), emails are sent from a template, and Custom runs your configured operation.

Note: Undeploying does not delete anything — it only takes the agent's live deployments offline until you redeploy it.

How Monthly Spend Is Counted

Each model call's cost is computed from its actual token usage at the model's catalog prices, and every metered invocation increments two running counters: the agent's monthly usage total (what the agent limit checks) and a per-user, per-agent usage record that carries the per-user limit alongside current and lifetime usage. The monthly window is a rolling ~30-day period rather than a calendar month: a daily job resets the agent's counter to zero 30 days after its current period started, whether or not the agent was active.

Alert rules are then evaluated on your Alert Interval against these totals (thresholds as a percentage of budget or an absolute amount), which is why an alert can fire a few minutes after the spend actually crossed the line rather than at the exact moment.

Managing Alerts Across Everything

A separate, workspace-level Alerts Manager lists alerts across all asset types in one table — with each alert's name, description, category, interval, and an active toggle — where you can create, edit, share, delete, and search them. Use the per-agent Spending tab to manage that agent's cost alerts; use the Alerts Manager to see and manage alerts everywhere.