Skip to main content

AgenticOps in one paragraph

You describe what you need in natural language; specialized AI agents do the work. Mention an agent with @name, optionally add a #tool to shape the output, then write the rest like you’d brief a colleague. The same syntax — @agent #tool [instruction] — works across every domain. No CLI, no scripting, no tool switching.
@alex #dashboard show me EC2 cost breakdown by instance type, last 30 days

Before you start

You need a workspace with at least one cloud connection — agents can’t return real results without one. If you haven’t set that up yet, follow the Setup guide first, then come back.

Pick your role

Each tab is one role, one goal, and one prompt you can paste today. Once you see the shape of the output, the follow-ups take you deeper.
Goal: Stop paying for cloud you don’t use. See where waste lives in under a minute.You need: AWS, Azure, or GCP connection. Agent: @alex.Quick win — find idle and oversized resources:
@alex find EC2 instances with <20% CPU utilization over the last 30 days, plus any unattached EBS volumes and unused Elastic IPs
Alex queries the cloud APIs, joins with utilization metrics, and returns a ranked list with projected monthly savings.Follow up:
@alex #recommend right-sizing for the top 5 by waste
@alex #dashboard cost trend by service for this quarter
@alex draft a reserved-instance plan for the stable workloads above

The syntax in 30 seconds

Every prompt has the same shape:
@agent #tool [instruction]
  • @agent — which specialist to ask. Anna, Alex, Oliver, Tony, or Kai. Anna is always available; the others activate when their matching connection is added.
  • #tool (optional) — what shape of output you want.
  • instruction — write it like a Slack message to a colleague.

Tool commands

CommandOutput
#dashboardInteractive visualization with charts
#reportDetailed analysis document
#recommendActionable recommendations
#alertSet up monitoring notifications
#chartData visualization

Tips that compound

  • Be specific upfront. “EC2 costs in us-east-1 for the last 30 days” beats “show me costs” — the agent spends less budget guessing what you meant.
  • Refine, don’t restart. Agents keep conversation context. “Drill into RDS for the items above” works.
  • Combine tools. Use #dashboard first to see the shape, then #recommend for the action on the same topic.
  • Let Anna coordinate. Anything that touches more than one domain — start with @anna and name the specialists. One consolidated answer beats five tabs.

The Agentic Loop — pick where to go next

Every CloudThinker module runs the same four-phase loop, continuously: Detect → Analyze → Resolve → Validate Agents detect signals from your environment, analyze them into a hypothesis or plan, resolve the work under your policy, then validate the outcome and feed it back into memory before the next iteration. The loop runs 24/7 — operational coverage is no longer bounded by who is on shift, who is awake, or who saw the alert first. Human-on-the-loop, not in every step. AgenticOps does not remove the human; it changes where the human’s attention is spent. Agents own detection, analysis, and the routine moves of resolution. Humans own judgment — approving changes that carry real risk, intervening on edge cases the policy doesn’t yet cover, and setting the goals that frame what the loop is working on. Every action is auditable, every change is reviewable, and the policy you write is the policy that runs. The four autonomy levelsnotify → suggest → approve → autonomous — are set per loop and gated by RBAC. Teams typically begin at suggest or approve for production-facing work, observe how the agent behaves in their environment, then raise the ceiling once a pattern has proven reliable. Over time the same loop demands less of your approval queue: more steps run autonomously, fewer require sign-off, and the human’s role narrows to defining what counts as risky rather than reviewing each individual change. Three module loops are covered end-to-end in the tutorial track. Pick the one that matches what you want to improve next:

Code Review Loop

Analyzes every PR with context from running infrastructure, past incidents, and team conventions. Validates the change with inline comments and patch suggestions. You merge; the agent catches what’s easy to miss. Best when production-bug escapes are the worry.

Cost Loop

Detects spend, utilization, and commitments across AWS, Azure, and GCP. Analyzes waste into remediation plans with projected savings. You approve changes; the agent finds them. Best when the cloud bill is climbing.

Incident Loop

Detects alerts, deploys, and topology shifts. Resolves by ranking hypotheses and running approved runbooks, then validates the fix held. You approve risky steps; the agent does the rest. Best when MTTR is your problem.