Usage Examples

These examples show workflows that you can do in a conversation with Claude. Any MCP-compatible AI assistant that connects to the DigiUsher MCP Server can do the same.

---

Example 1: Analyzing Cloud Costs by Service

Scenario: You are a FinOps practitioner. You prepare the monthly cost review, and you want to see the services with the highest spend of last month.

Prompt:

"Show me our cloud spending for February 2026, broken down by service. I want to see the top 10 services by cost."

What happens:

1. Claude calls list_organizations to find your organization ID.
2. Claude calls query_expense_data with these parameters:
   • start_date: "2026-02-01", end_date: "2026-02-28"
   • granularity: "total"
   • group_by: ["service_name"]
   • order_by: [{"field": "effective_cost", "direction": "desc"}]
   • limit: 10
3. If the results contain data source IDs, Claude calls get_expense_dimension_lookups to translate the IDs to names.

Expected output: A table of the 10 cloud services with the highest cost in February. The table gives the totals and the large changes against the period before.

Follow-up prompts that you can try:

• "Now filter this to only AWS accounts"
• "Break down the top service by region"
• "Show me the daily trend for EC2 costs this month"

---

Example 2: Investigating Cost Anomalies

Scenario: You get an alert about an unexpected cost increase, and you want to know the cause quickly.

Prompt:

"Are there any high-severity cost anomalies in my cloud accounts this month? Show me the details of the most impactful one."

What happens:

1. Claude calls list_organizations to find your organization ID.
2. Claude calls get_anomaly_summary with severity: ["HIGH"] and the date range of this month. This call returns the count.
3. Claude calls list_anomalies with the filter severity: ["HIGH"] to get the list.
4. Claude calls get_anomaly for the anomaly with the largest impact to get the full details and the impact metrics.

Expected output: The number of high-severity anomalies, and then the details of the anomaly with the largest impact. The details give the affected service, the region, the cost impact, the anomaly type (for example a spike or a pattern deviation), and the time when DigiUsher found it.

Follow-up prompts that you can try:

• "What about medium-severity anomalies?"
• "Show me all anomalies for the us-east-1 region"
• "Are there any pattern deviation anomalies?"

---

Example 3: Finding Cost Optimization Recommendations

Scenario: You are an engineering manager. You look for quick savings before the quarterly budget review.

Prompt:

"What open cost optimization recommendations do we have? Give me a summary of potential savings, then show the top 5 recommendations by savings amount."

What happens:

1. Claude calls list_organizations to find your organization ID.
2. Claude calls get_savings_summary with status: ["open"] to get the total possible savings.
3. Claude calls list_recommendations with status: ["open"], sorted by savings in descending order, with the limit 5.

Expected output: First a savings summary, with the possible monthly savings and annual savings of the open recommendations. Then a list of the 5 largest recommendations, with details such as the resource name, the current configuration, the recommended change, and the estimated savings.

Follow-up prompts that you can try:

• "Group the savings by scenario type"
• "Show me only the AWS rightsizing recommendations"
• "What recommendations have already been applied?"

---

Example 4: Tracking FinOps KPIs Over Time

Scenario: You prepare a FinOps maturity report. You must show the trend of the key metrics of the last quarter.

Prompt:

"How has our effective savings rate and compute commitment coverage trended over the last 3 months?"

What happens:

1. Claude calls list_organizations to find your organization ID.
2. Claude calls get_kpi_time_series with these parameters:
   • start_date: 3 months ago
   • end_date: today
   • kpi_ids: ["effective_savings_rate", "compute_commitment_coverage"]

Expected output: A time series with the daily or weekly values of both KPIs over the three months. Claude marks the trends and the large changes. Claude can give a table, or describe the trend in words.

Follow-up prompts that you can try:

• "What are all our current KPI values as of today?"
• "Show me the cost optimization index trend for the last 6 months"
• "How much commitment discount waste do we have?"

---

Example 5: Chargeback Analysis Across Teams

Scenario: You must report the cloud costs of each engineering team for the chargeback cycle of last month.

Prompt:

"Show me the cost allocation breakdown across teams for February 2026."

What happens:

1. Claude calls list_organizations to find your organization ID.
2. Claude calls get_chargeback_for_month with month: "2026-02-01" to get the chargeback data.
3. For a summary view, Claude calls get_cost_allocation_summary with start_date: "2026-02-01", end_date: "2026-02-28", and view: "chargeback". The view field is necessary. Use showback for the cost of each pool before the redistribution. Use chargeback for the cost after the redistribution.

Expected output: The costs of each team or pool. The result gives the total cost of each team, the percentage of the total spend, and the costs without an allocation.

Follow-up prompts that you can try:

• "Compare this to January's chargeback"
• "Show me Q1 2026 cost allocation with monthly granularity"
• "Which team had the largest cost increase month over month?"