Overview
Cloud Agents are scheduled AI workers that keep watch over recurring questions for your team. Instead of waiting for someone to remember to ask Atlas for an update, an agent can run on a schedule, apply its instructions, and produce intelligence cards for review.
This is useful for work that repeats: monitoring suppliers, checking shipment activity, watching a market or commodity, reviewing risk signals, or keeping an eye on a category of records. A good Cloud Agent turns a recurring research habit into a repeatable background workflow.
Cloud Agents are not meant to replace human judgment. They are meant to reduce the manual work of gathering signals. The agent can surface changes, summaries, and items worth reviewing, while your team decides what to trust, what to escalate, and what to do next.
The most important idea is scope. A broad instruction like "watch suppliers" will produce broad results. A strong instruction explains what to monitor, which entities or lanes matter, what kind of evidence is useful, and what a good intelligence card should contain.
How to access
Sign in to Atlas Verified.
Open Intelligence from the workspace.
Choose Cloud Agents.
Open an existing agent to review its latest cards and run history, or choose Add Cloud Agent to create one.
What you see on the Cloud Agents screen
System Agents are built-in agents provided by Atlas for common intelligence workflows.
Custom Agents are agents your team creates for your own monitoring needs.
Each agent row shows its name, description, type, enabled state, schedule, and most recent run status.
Opening an agent shows generated intelligence cards, run history, and the prompt that guides the agent.
The detail view lets you run an agent manually when you need an update before the next scheduled run.
What a good Cloud Agent prompt includes
The business goal: what the agent should help you notice or decide.
The scope: suppliers, shipments, commodities, routes, regions, or records to focus on.
The evidence standard: what sources or signals should matter, and what should be ignored.
The output shape: whether you want risk flags, summaries, recommended follow-up, or concise intelligence cards.
The cadence: how often the work should run based on how quickly the underlying situation changes.
Common uses
Monitor a high-priority supplier for certification, compliance, or operational changes.
Watch a commodity or trade lane for market movement, disruption, or unusual activity.
Review shipment-related intelligence on a recurring schedule.
Generate intelligence cards that teammates can review without rerunning the same research manually.
Keep a lightweight background watch on issues that are important but easy to forget during daily work.
Frequently asked questions
Do Cloud Agents take action automatically?
Cloud Agents generate intelligence for review. Treat their output as a starting point for human judgment, especially for compliance, supplier, or customer-facing decisions.
What is the difference between Atlas chat and a Cloud Agent?
Atlas chat is best when you have an immediate question. A Cloud Agent is best when the same question needs to be checked again and again over time.
Why did my agent produce weak results?
The most common cause is an instruction that is too broad. Edit the agent prompt so it names the business goal, the entities to monitor, and the kind of output you expect.