Skip to content

Cloud Cost Recommendations

import { Steps } from ‘@astrojs/starlight/components’;

Xplorr scans your connected cloud accounts and generates recommendations to reduce spend. Each recommendation includes an estimated monthly savings amount where the source provides one, and remediation steps where they are available.

The Recommendations page has three tabs:

  • Recommendations: everything found, to work through.
  • Approvals: requests to act on high value recommendations (see Approvals).
  • Realized savings: savings still to capture from open recommendations, next to savings measured in your bills after a recommendation is marked done.

Approvals and Realized savings were separate pages until 28 September 2026; their old addresses open the tabs.

SourceWhat it catches
Idle resources (Xplorr’s own check, AWS)EC2 instances averaging under 5% CPU over the past 7 days, and EBS volumes not attached to any instance. The check runs every 12 hours.
AWS Savings PlansSavings Plans purchase recommendations from AWS Cost Explorer
Azure AdvisorAzure Advisor cost recommendations for your subscriptions
Google Cloud RecommenderIdle VMs, VM machine type changes, idle persistent disks, idle Cloud SQL instances and committed use discounts

Each recommendation carries an estimated monthly savings figure when the source provides one. Provider recommendations are refreshed during the daily sync.

Every recommendation moves through a lifecycle:

Open → In progress → Applied
  • Open: newly generated, no action taken yet
  • In progress: someone on your team is working on it
  • Applied: the change has been made
  • Archived: set aside from Open or In progress; restore it to reopen it

On the Recommendations page in the console, an open recommendation has Start (moves it to In progress) and Archive; one in progress has Mark done (Applied) and Archive. Applied recommendations can be reopened, and archived ones restored.

Enterprise organizations can send an open recommendation for sign off before anyone acts on it. Click Request approval on the recommendation; it shows an Approval pending badge until an admin decides. See Approvals for how the approval itself is reviewed and decided.

Where a recommendation has remediation steps on file, the console shows them when you expand it: a numbered list of what to do, with a copyable CLI command where one applies. Recommendations without steps say so.

  1. Open Recommendations from the left sidebar and stay on the Open tab.

  2. Expand a recommendation to read its description and any remediation steps.

  3. Click Start to move it to In Progress. This signals to your team that someone is handling it.

  4. Make the change in your cloud account, using the remediation steps where they are provided.

  5. Come back and click Mark done to move it to Applied.

  • Running a remediation command on production without a maintenance window. Read each command before running it, and schedule disruptive changes.
  • Ignoring unattached volumes because each one is cheap. They add up, and nothing else will flag them.
  • Applying reserved instance recommendations without checking term length. A 3-year RI saves more per hour but locks you in. Make sure the workload will still exist in 3 years before committing.

How often are recommendations updated? Provider recommendations are refreshed during the daily sync. Xplorr’s own idle resource check runs every 12 hours.

Can I get recommendations outside the console? Yes. Ask Claude through the MCP server, whose get_recommendations tool filters by status, type, provider and account, or ask the Slack bot.

Do recommendations work for all providers? Yes. AWS, Azure and GCP all generate recommendations, from different sources: see the table above. Xplorr’s own idle resource check covers AWS only.

What if a recommendation is wrong? Archive it. If you see a pattern of incorrect recommendations, contact support.

  • Focus on the highest-savings recommendations first — sort by estimated savings and work down the list.
  • Archive what does not apply to you to keep the queue clean; archived recommendations can be restored.
  • Always verify remediation commands before running them in production.