> ## Documentation Index > Fetch the complete documentation index at: https://traceroot.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Get Started > Create your first alert and watch it evaluate This guide walks through creating an alert, checking it against the live preview, and reading its state in the Alerts list. ## Prerequisites You need: * A project with incoming traces. See [Tracing](/docs/tracing/get-started) if you don't have that yet. * A Slack workspace connected to your TraceRoot workspace, with a destination channel selected. Slack is the only place alerts are delivered. See [Slack delivery](/docs/alerts/slack) to connect it. * Member access to the project. Viewers can see alerts but cannot create or change them. ## Create an alert Open your project in [TraceRoot Cloud](https://app.traceroot.ai), select **Alerts** from the left nav, then click **New Alert**. The form has three sections on the left and a **Live preview** on the right. Alert form with the scope, condition and notify sections and the live preview chart showing the threshold line ### Set the scope Under **Scope**, use **Add filter** to limit which spans are measured, for example `Span kind is LLM`. Leave it empty to measure every span in the project. A metadata filter also needs a key. Half-filled filter rows are ignored. ### Set the condition **Condition** reads as one sentence: *aggregation* **of** *measure*, *operator* *threshold*, **over the last** *window*. For example, `p95` of **Latency** `>` `2000 ms` over the last `10m`. 1. Pick the **aggregation** and the **measure**. The aggregation list only offers what the selected measure supports: for example, **Count** takes only `count`, and **Latency** takes `avg`, `p95` and the other numeric aggregations. 2. Pick the **operator** and enter the **threshold**. 3. Pick the **window**: how far back each evaluation looks. The default is `10m`. The threshold field shows the measure's unit: `ms` for latency, `$` for cost and `tok/s` for total tokens per second. A rule for "p95 latency over 2 seconds" uses a threshold of `2000`. Token counts have no unit, and neither do the `count` and `uniq` aggregations. See the [rule reference](/docs/alerts/rule-reference#measures) for every unit. ### Choose how it notifies Under **Notify**: 1. **When a window has no data**: choose what an empty window means. The default, **Show no data, don't notify**, sends nothing. 2. **Renotify**: leave it **Off (alert only on transitions)**, or choose **Re-alert at a regular interval** to repeat the message while the rule stays in breach. 3. **Name**: how the alert identifies itself in Slack. The **Integration** row shows the Slack workspace and channel the alert will post to. Click **Create Alert**. New alerts are active immediately, and the first result appears within a minute. ## Check the live preview The **Live preview** charts the metric as you edit, so you can tune the rule before saving it. Over a short date range each point is one of the rule's windows; over a long range the points are coarser. * The **threshold line** is drawn across the chart and labelled with the trigger and its unit, for example `Alert > 2000 ms`. * The **shaded band** marks the side of the line that breaches. * The date range selector changes how much history the chart shows. It does not change the rule. If the chart says **No preview available for this metric yet**, the measure, aggregation and filters cannot be combined. Distinct users and distinct sessions cannot be combined with filters. ## Read the Alerts list Each row shows the rule's latest state: * **OK**: the latest value is within the threshold. * **Alert**: the latest value breaches the threshold. * **No Data**: the latest window had nothing to measure, or the rule has not run yet. * **Failing**: the last run failed and will retry next minute. * **Parked**: the rule's saved settings cannot be evaluated, so it has stopped. Edit and save the rule, or resume it, to start it again. * **Paused**: evaluation and notifications are stopped until you resume the rule. Click a badge to read why it shows what it does, including the reason a notification could not be delivered. ## Manage alerts Use the **Actions** menu on a row to pause or resume, edit, or delete a rule. * **Pause** stops evaluation and notifications. **Resume** starts the rule fresh: its state returns to **No Data** until the next evaluation. * **Edit** opens the same form. Changing the filters, aggregation, measure, operator, threshold, window or no-data setting also starts the rule fresh. Renaming the rule or changing renotify does not. * **Delete** removes the rule permanently. Starting a rule fresh closes any open alert without a recovery message in Slack. If the metric still breaches after an edit or a resume, the next evaluation announces it as a new alert. A project can hold up to **100 alerts**. Paused alerts count toward the limit. ## Next steps Every measure, unit, aggregation, filter and state. What each message contains and how to fix a failed delivery. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.