Real-time observability and anomaly detection

It explores the data sources connected to your workspace, including databases, warehouses, and synced SaaS tools, and can join across them in one analysis. It works through your models and skills, so it uses your metric definitions instead of guessing at them. On Basedash cloud it can also search the web for outside context, like a holiday behind a dip. Admins steer it with AI context in Settings → Insights: your goals, your priorities, and the metrics to focus on.

There's no fixed statistical model per metric. Each run, the agent explores your data the way an analyst would, compares recent periods with earlier ones, and looks for a change worth a post: a spike, a drop, a broken trend, or a milestone. When it finds one, it digs into the segments or events behind it and writes up what changed and why, with one chart. It validates the SQL before charting and checks that the time range it describes matches the data it actually has.

Admins choose the frequency in Settings → Insights: every day, every weekday, every week on a day you pick, or every month on a day you pick. Each scheduled run produces one post, so the frequency sets how many you get. You can also generate an insight by hand at any time, which is useful right after you change the AI context. Set the frequency to Never to stop scheduled runs and generate insights only on demand.

Every insight lands in the Insights feed in Basedash, where you can open its chart and keep exploring. In Settings → Insights → Notifications you can also send each one to email recipients and to a Slack channel, where the chart posts as an image. For a private Slack channel, invite the app with /invite @Basedash first. If you embed Basedash in your own product, the React SDK's BasedashInsights component shows the same feed there.

Insights are built to stay quiet: each run writes one post, and the agent reads recent insights first so it doesn't repeat a finding you've already seen. If posts still feel off-target, narrow the AI context to the goals and metrics you care about, or run insights weekly instead of daily. When you need an alert on an exact condition, like a failed payment above a set amount, use an automation with a data-change trigger. It runs only when your SQL query returns new rows.

Yes. The agent runs as a member of your workspace, with that member's data source access and permissions, so it can't query anything that member can't. Schemas, tables, or columns you've hidden from AI stay out of its context, and it uses the same governed models as the rest of Basedash. On Enterprise, its data access shows up in audit logs like any other query. Customer data is never used to train AI models.

Insights is included in the Startup and Enterprise plans, alongside automations, so there's no separate add-on to buy. The agent's work counts toward your plan's AI usage like any other analysis. You can start a free trial, connect a source, write your AI context, and generate your first insight by hand before you decide.