real quick backstory —
June 7th I checked my Copilot usage and it said $111 overage. no warning. no gate. just a number. that was the moment.
I’m not the type to just eat that and move on. I’m a builder. so I asked: why doesn’t a pre-execution cost check exist before a prompt fires? why does the AI ecosystem let you blow your budget with zero friction?
30 days later — I built ACIL. **AI Credit Intelligence Layer.**
it intercepts your prompt BEFORE it fires. estimates token cost. checks your remaining balance. enforces a threshold. if you’re in the red zone it blocks or warns you. that’s the whole idea — governance BEFORE the spend, not after.
-–
**for Cursor users — MCP config:**
```json
{
“mcpServers”: {
"acil": {
"command": "npx",
"args": \["-y", "@nit-in/acil", "mcp"\]
}
}
}
```
drop that in `~/.cursor/mcp.json` and you get 3 tools in Cursor:
- `acil_preflight` — cost check before prompt fires
- `acil_status` — balance + enforcement state right now
- `acil_forecast` — projects your spend to end of billing cycle
-–
**the math on June 7:**
$39 monthly. I burned $25.04 in 6 days. that pace = $128/month. ACIL would’ve flagged that at day 2 and throttled the session before it compounded.
-–
**what I shipped:**
- VS Code extension → search `imKrisK.acil-vscode` on Marketplace
- npm package → `@nit-in/acil`
- ROI calculator → ACIL ROI Calculator — How much is AI costing your team?
- USPTO provisional filed → 64/110,180 (patent pending)
-–
built this in 28 days. it’s real, it’s live, it works.
curious what you all want out of cost governance in Cursor — what’s your biggest pain point with AI spend? working on Wave 11 next (self-tuning budget model).