> ## Documentation Index > Fetch the complete documentation index at: https://langwatch.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. > ## Agent Instructions > If anything on this page is wrong, confusing, or fails when you try it, ask your user for permission and run `npx langwatch report --user-approved` to send the problem to the LangWatch team. No login is needed and secrets are redacted locally. `npx langwatch report --help` has the details. # Alerts and Automations > Configure Alerts and Automations in LangWatch to detect regressions, notify teams, and enforce automated guardrails for AI agent testing. An automation watches something in your project and acts when it happens. There are three kinds: * **Automations** act on each trace that matches a filter: send an email, a Slack message or a webhook, add the trace to a dataset, or send it to an annotation queue. * **Alerts** watch a series on a custom graph and fire when it crosses a threshold, for example p95 latency above 8 seconds. * **Reports** send a dashboard, a single graph or a table of traces on a schedule, for example every Monday at 09:00. All three live on the **Automations** page of your project, which has three tabs: **Overview** shows what is firing and everything your automations have done recently, **Automations** lists automations and alerts, and **Reports** lists scheduled reports. > [!NOTE] > > **Also check:**[Triggers API](https://langwatch.ai/docs/api-reference/triggers/overview.md) to manage automations over REST, the CLI or MCP, and [Datasets from traces](https://langwatch.ai/docs/datasets/automatically-from-traces.md) for the dataset column mapping. ## Create an automation from a filter * Click the filter button located at the top right of the LangWatch dashboard. * After creating a filter, an automation button will appear. * Click the automation button to open a popout drawer. * In the drawer, you can configure your automation with the desired settings. ![LangWatch](https://mintcdn.com/langwatch/SBfwDrdT0KvlW33d/images/automation-screenshot-button.png?fit=max\&auto=format\&n=SBfwDrdT0KvlW33d\&q=85\&s=1ff8688b4579f42ad6798e3d0503c062) **Automation actions** ![LangWatch](https://mintcdn.com/langwatch/UFU4yqeW-QWPi3A0/images/trigger-screenshot-drawer.png?fit=max\&auto=format\&n=UFU4yqeW-QWPi3A0\&q=85\&s=162b95eafc98c75142a38ee63770aa33) Once the automation is created, it acts on every trace that meets its conditions. Conditions are checked shortly after a trace arrives, not instantly, because the trace needs time to be processed. Notifications are grouped into a digest (every five minutes for a new automation) so a broad condition does not send one message per trace. A condition on an evaluation result names the **monitor** that ran the evaluation, not the evaluator. The dashboard picks this for you; over the API, see [Evaluation keys are monitor ids](https://langwatch.ai/docs/api-reference/triggers/overview.md#evaluation-keys-are-monitor-ids). **Automation settings** ![LangWatch](https://mintcdn.com/langwatch/UFU4yqeW-QWPi3A0/images/trigger-screenshot-settings.png?fit=max\&auto=format\&n=UFU4yqeW-QWPi3A0\&q=85\&s=d14c9d54f1ff19a27125e2ce8ce21241) ## Slack A Slack automation posts through a **Slack connection**. Connections live under **Settings → Integrations → Slack**, where each one is listed with its name, kind, scope and how many automations use it. Add one with **Add Slack connection**, or with **New Slack connection** while setting up an automation's Slack delivery; the new connection comes back selected. A connection is one of two kinds: * **Bot**: a Slack app bot token. Pick a channel for each automation, and the app posts there with every block of the message. Invite the app to the channel first. * **Incoming webhook**: one webhook URL, which posts to the one channel Slack tied it to. Charts and tables are sent as text. Each connection has a scope: * **Organization**: every project in the organization can use it. Adding, editing or deleting one needs permission to manage the organization. * **Project**: only that project can use it. Adding, editing or deleting one needs permission to update the project. A token or URL is stored once per organization: adding a connection with a secret another connection already holds is refused, naming that connection. Replacing a connection's token or URL takes effect for every automation that uses it, with no automation edited. Deleting a connection that active automations use asks you to confirm how many stop delivering. > [!NOTE] > > Automations used to carry their own webhook URL or bot token, and a project had one Slack integration. Each of those secrets is now a connection, and every automation points at the connection holding its secret: automations that shared a secret within one project share a project connection, and a secret shared across projects became an organization connection. You do not need to re-enter any secret. ## Alerts and reports An **alert** is set up on a custom graph: pick the series to watch, the comparison (above, below, equal), the threshold and the window (1 minute to 1 day). It fires when the series crosses the threshold and records when it recovers. A **report** renders a dashboard, one graph, or a table of the top traces for a filter, and sends it by email or Slack on a cron schedule in the timezone you choose. A report sends at most once every 15 minutes. Both can also be created with `langwatch trigger create --graph-alert` and `--report`; see the [Triggers API](https://langwatch.ai/docs/api-reference/triggers/overview.md#alerts-on-a-graph). ## Pause, test and history * **Pause** an automation to stop it acting without deleting it, and resume it later. A paused report stops sending. * **Test** sends the automation's message to its saved destination so you can check it arrives. A test is not recorded as a fire. * **History**, in an automation's details, lists every fire, newest first; **Overview** shows the recent fires across all of them. The same history is available with `GET /api/triggers/{id}/fires`, `langwatch trigger fires ` and the MCP tool `platform_list_trigger_fires`, each paged by cursor. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.