In early March, we launched the CloudZero AI Hub and the CloudZero Claude Code plugin, giving customers a direct line to their cloud and AI cost data through natural language.
Early adopters and power users have already jumped in, using the plugin to investigate cost spikes, close commitment gaps, and get to cost-per unit metrics that used to take days to pull together.
What we’ve noticed over the past few weeks is pretty consistent (and predictable). The teams getting the most significant value from the MCP integrations in our AI Hub are the ones who came in knowing exactly what they wanted to ask.
So we built the AI Prompt Catalog to help close that gap for everyone else.
Today we’re releasing 46 pre-built prompts organized by maturity level, role, and use case. Whether you’re a FinOps analyst who’s been running cost analysis for years or a finance executive jumping into CloudZero for the first time, there’s an easy starting point here for you. The prompts cover the full scope of what CloudZero tracks: cloud infrastructure, SaaS and PaaS spend, and AI services.
Transparently, we started working on this as soon as we released the AI Hub because we knew people would be coming at this from very different starting points. Some companies have mature, fully-staffed FinOps teams and just need to get their hands on the right tools. Others are earlier in the process and still figuring out where to begin. We wanted to make sure neither group had to slow down to think about how to build the first set of prompts that would be valuable to them.
AI cost management is a key area we knew would probably need some additional support. It’s new territory for most teams, and flying blind on it is more common than people want to admit. We’ve built several prompts to help address that gap and answer questions like, “Which of our AI models and endpoints are the most expensive to run, and how is that trending over time?” Most teams can’t answer that today. We wanted to make it easier.
What we heard from early adopters
The customers getting the most value from all the integrations in our AI Hub tend to share a few traits. They’ve already mapped their cost allocation model. They know which services drive the most spend. And they came to the table with specific, well-formed questions rather than open-ended ones.
We highlighted this pattern during our webinar last week, where some of our own experts walked through exactly how they’re using the plugin in their day-to-day workflows. The throughline: specificity wins. The more context you give Claude about what you’re looking for, the more useful the output.
This tracks with what the research shows. A Harvard/BCG study of 758 knowledge workers found structured prompt guidance made them 25% faster, with 40% higher-quality output. The difference isn’t always in the model, it often comes from the question itself.
That’s what these prompts are. They’re the specific, well-formed questions that are already tested against real CloudZero data, ready to copy and run.
Research Report
FinOps In The AI Era: A Critical Recalibration
What 475 executives told us about AI and cloud efficiency.
How the catalog is organized
Every prompt is tagged by maturity level, persona, and use case so you can filter to exactly what’s relevant for your team and where you are in your FinOps practice:
- Crawl prompts cover basic visibility: spend by team, top cost drivers, budget variance, AI and ML service spend. Good starting points for anyone new to the plugin or to FinOps tooling in general.
- Walk prompts get into optimization and accountability: commitment gap analysis, rightsizing opportunities, unallocated spend, quarter-close reconciliation.
- Run prompts tackle advanced analysis and multi-tool workflows that pull in data from GitHub, Datadog, Jira, PagerDuty, and Slack alongside CloudZero data.
Three prompts worth highlighting:
1. Budget Variance by Team (Crawl — Finance, FinOps)
“Are we over or under budget this month, and which teams are driving the variance?”
Simple question, high-value output. Finance teams spend real time pulling this together manually every month by cross-referencing budget files, cost reports, and team allocations. This prompt surfaces the answer in seconds and points directly to where to look next. It’s a good first prompt for anyone on the finance side who hasn’t used the plugin yet.
2. Cost Per Customer by Feature (Walk — Product, Finance, FinOps)
“What is our average infrastructure cost per feature over the past quarter, ranked from most to least expensive? Also show total cost by customer over the same period. Flag what percentage of total spend is unallocated to a specific feature, since that limits the precision of per-feature unit economics.”
This is one of the prompts we’re most proud of, because it’s asking a question most cloud cost tools simply can’t answer. Engineering wants to know which features are expensive to run. Product wants to know if the margin on a feature justifies its roadmap priority. Finance wants the fully-loaded number. This prompt sits at the intersection of all three, and the output isn’t just a cost report. It flags how much spend is unallocated, which tells you exactly how much to trust the per-feature numbers. That kind of built-in precision check is what separates a useful answer from a misleading one.
3. Cost Spike Root Cause via GitHub (Run — Engineering, FinOps, DevOps)
“There was a significant cloud cost spike in the past 7 days. Using CloudZero, identify which services, teams, or cost dimensions show the sharpest increase and pinpoint the exact day the spike began. Then use GitHub to pull all commits, merges, and deployments that went out in the 48 hours before the spike started…”
This is where the multi-tool capability of Claude Code gets interesting. Cost spikes don’t happen in a vacuum. They almost always correlate with something that changed in the codebase. This prompt pulls CloudZero cost data and GitHub deployment history simultaneously, overlays the timelines, and produces a root cause hypothesis. What used to take an engineer and a FinOps analyst comparing notes across two different tools now happens in a single query.
Getting started
The full catalog is at prompt.cloudzero.com. Every prompt is filterable by maturity level, persona, and use case, and copyable in one click.
If you’re a current CloudZero customer, you can start using these with the Claude Code plugin today. If you’re evaluating CloudZero and want to see the plugin in action, contact your account team or request a demo.
FinOps In The AI Era: A Critical Recalibration
What 475 executives told us about AI and cloud efficiency.
