Works with the AI Agents you already use
Cost Per User
Before you launch, or before you set a price, you want one answer: **will each user make me money, or cost me money?** Cost Per User reads your code and works it out.
$15
Cost Per User
Example session with this skill installed
What will my recipe app (recipebox) cost to run at 1,000 users? I charge $6 a month and about 15% of users pay. Is it making money?
- Read your context and instructions
- Compiled the cost per user
The short answer (sample app: "recipebox", made up for this example)
- 100 users: about $176 a month. LOSING MONEY
- 1,000 users: about $1,373 a month, $1.37 per user. LOSING MONEY
- 10,000 users: about $13,148 a month. LOSING MONEY
One line of code is 85% of the bill: the meal plan page asks the AI for a brand new plan every time someone opens it. At $6 a month with 15% paying, income is $900 at 1,000 users, so you lose about $473 a month. Store the plan and rebuild it once a week, and the same 1,000 users cost about $287 a month with a 68% margin.
| Line | 100 users | 1,000 users | 10,000 users |
|---|---|---|---|
| Vercel Pro | $20 | $20 | $20 |
| Supabase Pro | $25 | $25 | $25 |
| AI recipe import | $7 | $73 | $732 |
| AI meal plan, every page load | $116 | $1,164 | $11,640 |
| Resend (email) | $0 | $20 | $20 |
| Stripe fees | $7 | $71 | $711 |
| Total per month | $176 | $1,373 | $13,148 |
| Income | $90 | $900 | $9,000 |
Cost bomb 1: a new AI meal plan on every page load (app/plan/page.tsx:14). The page calls the AI on each request and never saves the answer, so 60 visits a month means 60 plans when one a week would do. Costs $1,164 a month at 1,000 users; about $78 after the fix.
Fix prompt: Store the generated plan in a meal_plans table with the user id and the week. On page load, read the stored plan; only call the AI when there is no plan for this week or the user presses "New plan". Keep the prompt and model the same.
Next steps
- Paste the fix prompt into your AI builder, then re-run Cost Per User to see the new bill
- Set a monthly spend limit on your AI provider
Connects securely to your tools. The creator never sees your data.
About this skill
You built an app with AI. Before you launch, or before you set a price, you want one answer: will each user make me money, or cost me money? Cost Per User reads your code and works it out.
Ask "what will this app cost me at 1,000 users?" and you get one report:
- The monthly bill at 100, 1,000 and 10,000 users, line by line, with cost per user, profit, margin and a plain verdict (healthy, thin, at risk, losing money)
- Every paid service your app uses, each with the file and line: hosting, database, login, storage, email, texts, payments, AI models and other APIs, including ones called with plain
fetchand jobs hidden in GitHub Actions - Cost bombs: the lines that make the bill grow much faster than your users. An AI call on every keystroke. A timer that calls the AI for every user every five minutes. Texts that bill as four messages each because of one emoji
- A copy-paste fix prompt for each bomb, and the bill after the fixes
- Price cliffs: the user counts where a free plan runs out
- Every number labelled: from your code, from the dated price sheet, told by you, or assumed. Change one in
cost-model.json, re-run, and get a new bill
Why a calculator, not a guess The skill ships a small Python calculator (standard library only, no network). The AI counts what each action costs from your code; the script does the sums. Totals are computed, not invented.
Prices included A dated price sheet for 25+ common services and current AI models (OpenAI, Anthropic, Google, Vercel, Supabase, Neon, Firebase, Clerk, Resend, Twilio, Stripe, Cloudflare and more), with token-counting rules. If your AI tool has web access it re-checks anything older than 60 days; without web access it uses the sheet and says so.
Safe
- Read-only: it writes only
cost-report.mdandcost-model.json - Never calls a paid service and never prints your keys
- Says "No cost bombs found" when the code is careful, instead of padding the list
Tested in Claude Code on five sample apps with planted cost problems (Next.js, Express, a 60-file marketplace with nine services) and a well-built control app, 33 runs in all. With the skill, every report found every paid service and every planted cost bomb, made the right profit call, and landed within 30% of the hand-made answer key in 12 of 13 runs. Without web access, the same model without the skill was within 30% in 6 of 10 runs and was more than 2x off in 3, mostly because it had to guess prices for current AI models; the skill's dated price sheet covers them. Honest note: when the model could search the web, it did about as well without the skill. What the skill adds there is the method and the format: the same labelled report every time, a re-runnable model file, price cliffs and fix prompts.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
Requires a local Python environment for exact math. Cannot access live billing dashboards or private pricing tiers not in the reference sheet. Estimates only.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 day ago
- Software
Python 3 (for the bundled calculator)