Once your assistant is connected to the CloudFix MCP server, you don’t need to know which tool to call. Ask the question the way you would ask a colleague, and your assistant picks the tool. The prompts below are a starting point, grouped by who tends to ask them, with the CloudFix tool that answers each one.
Every tool reads. None of these prompts can change your AWS account: when you decide to act, you approve the fix in CloudFix and it runs in your account.
Where is the money?#
| Ask |
Tool your assistant uses |
| “What are my biggest AWS savings opportunities right now?” |
get_recommendations_summary |
| “List my ten largest recommendations by annual savings.” |
list_recommendations |
| “Which savings can CloudFix fix automatically, and which need a maintenance window?” |
get_recommendations_summary |
| “Show me every open recommendation in account 123456789012.” |
list_recommendations |
| “What has CloudFix found in eu-west-1 this month?” |
list_recommendations |
| “Break down potential savings by organizational unit.” |
get_recommendations_summary |
Engineers, in the editor#
| Ask |
Tool your assistant uses |
| “What does an m7g.2xlarge cost in us-east-1, on demand and on a 1-year No Upfront plan?” |
get_ec2_price |
| “Find the cheapest Graviton instance with at least 8 vCPUs and 32 GB of memory in us-west-2.” |
find_ec2_instances |
| “Is there a cheaper instance family with the same vCPU and memory as the one in this Terraform file?” |
find_ec2_instances |
| “Has CloudFix flagged anything on the resource IDs in this module?” |
list_recommendations |
| “What exactly would change on this volume, and what does it save?” |
get_recommendation_details |
| Ask |
Tool your assistant uses |
| “What is waiting in Manual Approval, sorted by annual savings?” |
list_recommendations |
| “Which fixes failed, and in which accounts?” |
list_recommendations |
| “What has been scheduled but not run yet?” |
list_recommendations |
| “How much have we saved from completed fixes so far?” |
get_recommendations_summary |
| “Show me the evidence behind this rightsizing recommendation before I approve it.” |
get_recommendation_report |
| “Which recommendations have I rejected, and what were they worth?” |
list_recommendations |
CTOs and finance: the update#
| Ask |
Tool your assistant uses |
| “Summarize our AWS savings for this week’s leadership update in five bullet points.” |
get_recommendations_summary |
| “Compare potential, completed and realized savings, and tell me what is holding up the rest.” |
get_recommendations_summary, list_recommendations |
| “Turn the top five open recommendations into a one-page brief my team can act on.” |
list_recommendations, get_recommendation_details |
Tips that get better answers#
- Name the scope. An account ID, region, organizational unit or status narrows the answer and keeps it fast.
- Ask for the evidence. Every recommendation has an analysis report behind it. Ask for it before you approve a change.
- Combine it with what your assistant already knows. Your assistant can put CloudFix’s findings next to your code, your tickets or your notes, which is where the useful answers usually come from.
- Keep the change in CloudFix. Your assistant can tell you what to approve. The approval, the schedule and the fix itself happen in CloudFix, with a full audit trail in your account.
Not connected yet? Set up the CloudFix MCP server in three steps, or get a free savings assessment to give your assistant something to read.
