AS
AI startup“We were flying blind on LLM costs for months. CostCap showed us our summarizer alias was eating 70% of our bill.”
4-person team
Real-time AI spend tracking
Stop discovering your AI bill three weeks after the sprint.
Track every LLM call in real time. Set hard caps per environment. Get alerted before you hit the ceiling, not after.
View setup$10/mo, cancel anytime
Built for
OpenAIAnthropicGoogleMistralCohere
Monthly AI spend
$428.1962% of cap
projected
$690
OpenAI
$212
Anthropic
$129
Google
$46
Mistral
$41
prod
80% alert
staging
$32.16
summarizer
70% of bill
Alert: prod is at 80% of cap. Next notification at 100%.
What teams are saying
OpenAI
FD
Freelance dev“I set a $50/mo cap on my dev environment and stopped worrying about runaway test loops racking up bills.”
Solo product builder
AG
AI agency“Finally able to bill clients accurately. We track each project under its own alias and export the numbers directly.”
Client AI work
What you get
Spend control that looks like a product, not a billing autopsy.
Real-time spend tracking
See every LLM call by model, provider, alias, environment, and project. Token counts, estimates, and duration are captured server-side.
Environment-level caps
Set monthly limits per environment. Dev, staging, and prod get separate budgets. Alerts fire at 80% and 100% of each cap.
Provider breakdown
See spend across OpenAI, Anthropic, Google, Mistral, and more in one view. Know which provider is costing you the most.
One-line SDK
Wrap any LLM call with vibe.track(). Fire-and-forget logging keeps your response path clean.
API key management
Generate and revoke keys. Each key gets its own usage history. Hash-stored and never readable after creation.
Server-side cost recomputation
We never trust client-sent estimates. Costs are recomputed from our pricing table for 40+ models across 5 providers.
Good for
Stop letting experiments, staging, and client work share one mystery bill.
Teams shipping AI features
Know which features are expensive before they go to prod. Track spend by alias so each team owns its costs.
per-feature spendalias breakdownmodel comparison
Budget-conscious projects
Set a monthly ceiling and get warned before you hit it. No more discovering you've blown the budget at month-end.
hard caps80% alertsprojected EOL spend
Multi-environment setups
Dev team can't accidentally run up a prod-scale bill. Each environment gets its own independent budget.
dev capstaging capprod cap
Client work
Bill clients accurately. Track spend by alias per client to match invoices to actual LLM usage.
per-client trackingalias-based billingexport spend
How it works
Wrap the call. Track the spend.
Add CostCap around your LLM call, tag the environment, and see spend update instantly in your dashboard.
Install the SDK
Add the package and connect your API key.
Wrap your LLM call
Use vibe.track() without changing your response flow.
Track spend and caps
Monitor cost by model, provider, alias, project, and environment.
track-summarizer.ts
const result = await vibe.track(
"summarizer",
() => openai.responses.create({
model: "gpt-4.1-mini",
input: prompt,
}),
{ environment: "prod", project: "app" }
);Pricing
Simple pricing. No surprises.
One plan for teams that need real visibility before the provider invoice lands.
$10/ month
Everything included. No usage tiers.
- Up to 5 API keys
- Unlimited LLM calls tracked
- Real-time spend dashboard
- Environment-level spend caps
- Alerts at 80% and 100% of cap
- Provider breakdown (OpenAI, Anthropic, etc.)
- Spend by alias and environment
- API key management
Cancel anytime. No contracts.
Ready to stop guessing?
One line of code. Real-time visibility.
Add CostCap to your LLM calls and see exactly what your AI features cost by alias, environment, provider, and project.
FAQ
Details people check before trusting a spend tracker.
What counts as a tracked call?
Any LLM call wrapped with vibe.track(). Supports OpenAI, Anthropic, Google, Mistral, Cohere, and more. Unknown models still log token counts and request counts — cost will show as $0.
Does tracking add latency?
No. CostCap uses a fire-and-forget pattern. Your LLM call completes and returns first. Usage is logged asynchronously — zero latency added to your response times.
Can I set different caps for dev and prod?
Yes. Caps are per-environment. Set a low limit for dev and staging, a higher one for prod. Each environment tracks and alerts independently.
What triggers an alert?
Alerts fire at 80% and 100% of your monthly cap per environment. You can configure an alert email per cap. No alert email? The cap still enforces — you just won't get a notification.
How are costs calculated?
Server-side, from our pricing table. We never trust client-sent estimates. Covers 40+ models across 5 providers. Pricing is updated as providers change their rates.
How do I deploy?
CostCap runs on your own infrastructure. Deploy to Vercel, Railway, or Coolify. Set your DATABASE_URL (Supabase or Postgres), Stripe keys, and BETTER_AUTH_SECRET, then run prisma migrate deploy.