Prompt Guide
Prompt templates
Mavvrik MCP works best when the question includes a clear cost area, time period, and level of detail. A useful prompt usually states what needs to be reviewed, the period to review, how the result should be grouped, and whether the answer should be summarized or shown as a table.
A simple prompt pattern is:
“Show [cost area] for [time period], grouped by [dimension], limited to [top N], with [summary/table/drivers].”
Example:
“What drove cloud spend last week? Show the top 5 services by cost increase and summarize the main drivers.”
Cloud cost prompts
Use these prompts to review spend and identify cost drivers.
Use case | Prompt example |
|---|---|
Review recent spend | “Summarize cloud spend for the last 7 days.” |
Find top cost drivers | “Show the top 5 services by cloud cost this month.” |
Compare spend by period | “Compare cloud spend this week with the previous week and show the largest changes.” |
Review a specific service | “Show EKS spend for the last 7 days and identify the main cost drivers.” |
Review by account or scope | “Show cloud spend by account for this month and highlight the top 5 accounts.” |
Cost variance prompts
Use these prompts to understand why spend changed between two periods.
Use case | Prompt example |
|---|---|
Explain a cost increase | “Why did cloud spend increase compared with the previous week?” |
Investigate a specific service | “Why did EKS spend increase compared with the previous week?” |
Identify largest movers | “Show the top cost variances from the last 7 days compared with the previous 7 days.” |
Summarize positive and negative changes | “Summarize the biggest cost increases and decreases this month compared with last month.” |
Focus on likely drivers | “Show the services that contributed most to the cost increase last week.” |
Anomaly prompts
Use these prompts to review unusual or unexpected cost patterns.
Use case | Prompt example |
|---|---|
Review recent anomalies | “Are there any cost anomalies in the last 7 days?” |
Investigate a specific anomaly | “Explain the largest cost anomaly from last week.” |
Focus by service | “Show cost anomalies for EKS in the last 14 days.” |
Prioritize investigation | “List the most important anomalies from the last 7 days and explain why they matter.” |
Summarize anomaly impact | “Summarize recent anomalies and show their estimated cost impact.” |
Recommendation prompts
Use these prompts to review available savings opportunities.
Use case | Prompt example |
|---|---|
Review open recommendations | “What savings recommendations are currently open?” |
Prioritize recommendations | “Show the top savings recommendations by estimated savings.” |
Focus on a service | “Show open recommendations related to EKS.” |
Summarize action areas | “Summarize the main optimization opportunities available right now.” |
Compare recommendation types | “Group open savings recommendations by type and show the estimated savings for each group.” |
Follow-up prompts
Follow-up prompts are useful after an initial answer. They help narrow the analysis without starting over.
Examples:
“Break that down by service.”
“Show only the top 5 drivers.”
“Compare the same view with the previous week.”
“Focus only on EKS.”
“Summarize this as three key findings.”
“Show the recommendations related to these cost drivers.”
Prompting best practices
Mavvrik MCP works best with focused questions. A prompt should be specific enough to guide the analysis, but not so broad that the response has to process too much data at once.
Start with a recent and bounded time period. For most cost investigation workflows, the last 7 days, last 14 days, current month, or previous month is a good starting point. Very large time ranges, such as several months of detailed data across many services or accounts, can produce slower or less focused responses. For longer periods, ask for a summary first, then follow up with a narrower question.
Instead of | Use |
|---|---|
“Analyze all cloud costs.” | “Summarize cloud spend for the last 7 days and show the top 5 services by cost.” |
“Why did cost change?” | “Why did cloud spend increase last week compared with the previous week?” |
“Show everything about anomalies.” | “Show the most important cost anomalies from the last 7 days.” |
“Analyze the last 12 months in detail.” | “Summarize cloud spend trends for the last 3 months, then show the top 5 changes.” |
Specify the type of analysis needed. Mavvrik MCP can answer different kinds of cost questions, so the prompt should make the intent clear. For example, ask whether the goal is to review total spend, explain a variance, find anomalies, or identify recommendations.
Use:
“What drove cloud spend last week?”
Instead of:
“What happened with cloud cost?”
Ask for a useful level of detail. Top-N prompts usually work better than open-ended prompts because they keep the answer focused.
Use:
“Show the top 5 services by cost increase last week.”
Instead of:
“Show everything that changed last week.”
Add a dimension when needed. Cost questions are easier to answer when the prompt says how the result should be grouped or filtered. Useful dimensions may include service, account, region, or another available Mavvrik filter.
Examples:
“Show cloud spend by service for this month.”
“Show cloud spend by account for the last 7 days.”
“Show cost anomalies by account for the last 14 days.”
Avoid asking for too much depth in one prompt. Detailed multi-month breakdowns, large tables, and broad “analyze everything” requests can make responses harder to read and less useful. A better pattern is to start with a summary, then ask follow-up questions for the specific area that needs investigation.
Example investigation pattern:
“Summarize the biggest cloud cost changes this month.”
“Which services contributed most to the increase?”
“Show recommendations related to those services.”
Use clear comparison periods for variance questions. Cost variance prompts should define both periods when possible.
Use:
“Compare cloud spend for the last 7 days with the previous 7 days.”
Instead of:
“Why did cost change recently?”
Keep the question inside current MCP coverage. The current release supports Cloud Cost, Cost Variances, Recommendations, and Anomalies. Questions outside these areas may return limited results or may not be supported yet.
Use:
“What savings recommendations are open?”
Instead of:
“Create a full budget forecast and commit plan for next year.”
Ask for the response format when it matters. A prompt can request a short summary, table, ranked list, or investigation-style explanation.
Examples:
“Show the answer as a table.”
“Summarize in three bullets.”
“Rank the drivers by cost impact.”
“Explain the likely cause in plain language.”
Prompt quality improves the usefulness of the response, but responses still depend on the data available in the connected Mavvrik tenant and the authenticated user’s access. If the requested data is unavailable or outside the user’s permissions, the response may be limited.