Programming · Blog · 🇮🇳 India · via agentusecases.ai
A leaner MCP server finds wasted AWS spend in fewer calls
A cloud architect rebuilt a 15-tool MCP server with just 4 tools, and the agent now answers where an AWS account is wasting money in 2 calls instead of 7.
- Profession
- Developer, DevOps
- Tools
- Kiro, MCP, AWS
- Source
- dev.to ↗
- Country
- 🇮🇳 India
- Found via
- agentusecases.ai (Alex Finn) ↗
- Report date
- October 26, 2026
- Verified on
- October 29, 2026
See more: AI for developers
What they did
Sarvar Nadaf built a read-only MCP server so the Kiro agent could answer one question: where am I wasting money in this AWS account? The first version had 15 tools, each mirroring a single API call. In the second, he merged them into 4 tools that each deliver a complete result, with fixed-choice parameters, short responses of five fields per finding and clear error messages. Both versions are published on GitHub for comparison.
How it went
On the same account with the same question, the agent went from 7 calls to 2, and the text of the tool definitions shrank by 88%. Both versions found the same three items: an unattached disk volume (about US$ 0.80 a month), an orphaned snapshot (about US$ 0.40) and an idle Elastic IP, which the agent labeled "not verified" because it could not find the price. The author admits that in such a small demo the context savings were small (from 1.1% to 0.8%).
Good to know
The context savings really matter in setups with dozens of tools loaded. Give the agent read-only access to the account and review the findings before deleting any resource.
The prompt
I have an MCP server with these tools: [list of tools with descriptions]. Users usually ask for [most common tasks]. Propose a version with at most [number] tools that each complete a whole task in one call, with fixed-choice parameters where possible, short responses by default and an option to request full detail. Show the new signature of each tool with a description of up to two lines, and keep everything read-only.
Replace what is in [brackets]. Prompt written by us from the account, not a literal copy of the source. Do not paste other people’s personal data (privacy laws such as LGPD and GDPR apply).
Summary in our own words. Results as stated in the original account; we have not tested or audited them. Read the source ↗ Case found via agentusecases.ai (Alex Finn); text and prompt rewritten by us. Found a mistake?
Similar cases
01Letting the agent deploy and read production errors on its own
The founder of a deployment platform explains how they designed an MCP server to take the human out of "copy and paste the log".
02Livelo engineers report up to 50% higher productivity with AI agents
The Brazilian loyalty program brought Kiro and Amazon Q Developer to its team of about 240 engineers, with metrics and governance from day one.
03Launching a habit app solo and doing the marketing with AI
A dev created Kada, a habit app with no account and no ads, and used AI for marketing.
04Building an "internet TV" in 6 days with a team of agents
A dev created Crom TV, with 24-hour channels of YouTube videos, coordinating several AI agents.
05Asking AI for a specific algorithm for a research project
A developer skeptical of LLMs used Gemini to generate random quantum circuits in the QCOP project.
06Teaching an agent to use your bank's API
A developer created an open source skill so Codex and Claude Code can integrate with the Banco Inter API.
07Set up a 100% local AI coding environment
A junior developer integrated local models into VS Code with Ollama and Cline, with no paid APIs.
08Build a World Cup tracking website with AI help
An engineer built Central da Copa 2026 and explains why it wasn't "just asking the AI to do it."