Agents

Agents are scheduled, workspace-scoped automations powered by the platform assistant. Each agent runs on a cron-like schedule, queries your catalogue through the same scoped tools the chat assistant uses, and produces a run output — a report, an alert, or a notification.

How an agent runs

• Scoped to a workspace. An agent executes with its owner's access scope and role, so it sees exactly the assets that user can — no more.
• Tool-calling. It calls catalogue tools (search assets, read a control module, pull history, render a chart) to gather data, then summarizes or acts.
• Scheduled. A schedule (e.g. hourly, daily) triggers each run; runs also can be triggered manually.
• Run output. Every run records its output and is visible in the UI, so you can review what the agent saw and concluded.

Templates & the marketplace

• Templates give you ready-made agents (e.g. a shift summary, an anomaly digest) to start from.
• The marketplace lets you publish an agent and install others' as a disabled copy you then review and enable, or import one from JSON.

What an agent can do

Beyond reporting, an agent can raise an alert — which fans out through the notification channels (in-app and, if configured, email). This makes agents a way to turn catalogue conditions into proactive signals, not just dashboards you have to watch.

Prerequisites

Agents require the assistant to be enabled (an LLM connection configured in Settings, and the platform assistant master switch on). See Platform functionality for the assistant and notification building blocks, and note that the agents feature is license-gated like other premium features.