Never get a surprise AI bill again.
BurnGuard sits between your app and every AI provider — metering each request in real time and slamming the door the instant you hit your budget.
brew install burnguard · one line, zero SDK changes
$47.23
this month
1,847
2.4M in · 892K out
$47.23 used · 63%
click around · drag the cards · edit the budget · it's live
Watch every request get priced in real time.
Each call flows through the meter and lands on your dashboard the moment it completes. When the next request would cross your cap, BurnGuard rejects it — before the charge ever happens.
A developer left a loop running. The bill hit $30,141.
No cap. No alert. No kill switch. By the time the invoice email arrived, the damage was done — and the provider doesn't issue refunds for “working as intended.” This happens every week, to careful engineers.
- $30,141
- Runaway loop, one weekend
- 2,400,000
- Requests before anyone noticed
- $0
- Refund from the provider
Three lines between you and a five-figure mistake.
No SDK rewrite. No vendor lock-in. It's a proxy — it just watches the money.
Point your SDK at BurnGuard
Change one line — your base URL. BurnGuard runs as a local proxy, so your keys never leave your machine.
base_url = http://localhost:4000
Every request is metered live
Tokens and cost are computed per request across Anthropic, OpenAI, and Google(In View) — streamed to your dashboard in real-time.
SSE-safe · sub-millisecond overhead
Over budget? Blocked instantly
Set a monthly cap. The moment a request would cross it, BurnGuard rejects it and pings you — before you're charged.
429 budget_exceeded
Everything you need to sleep at night.
A firewall for your API spend — with the visibility to back it up.
Real-time cost tracking
Every request priced the instant it completes. Watch spend move as it happens, not at month-end.
Hard budget enforcement
Set a cap and mean it. Requests over budget are rejected at the proxy — the charge never lands.
Multi-provider, one view
Anthropic, OpenAI, and Google(In View) spend unified in a single dashboard with per-model breakdowns.
SSE streaming, untouched
Streaming responses pass straight through with usage captured on the fly. No buffering, no lag.
Alerts before it hurts
Slack and Discord webhooks fire at 50%, 80%, and 100% of budget so nothing sneaks up on you.
Keys never leave your box
The proxy runs locally. Your provider keys stay on your machine — BurnGuard only sees the meter.
Costs less than one runaway request
Start free. Upgrade when the budget guardrails become non-negotiable.
Free
For solo builders and side projects.
- 1 sync token
- 7-day spend history
- Real-time tracking
- Single provider
- Community support
Pro
For teams shipping on AI in production.
billed annually
Protect Your Wallet- Unlimited sync tokens
- Unlimited history
- Hard budget enforcement
- All providers unified
- Slack + Discord alerts
- Priority support
Trusted by developers who've felt the burn
“We caught a runaway agent at 3am — BurnGuard blocked it at $12 instead of $12,000. It paid for itself the first night.”
“Finally one dashboard for Anthropic and OpenAI spend. The hard budget cap is the feature I didn't know I was desperate for.”
“Setup was genuinely one line. It's been running in prod for months and I've stopped checking billing dashboards.”
Questions, answered.
It runs as a local proxy. Before forwarding a call, it checks the projected cost against your budget. If the request would push you over, it never reaches the provider — you get a clean 429 back instead of a charge.
No. The proxy runs locally and holds your provider keys. BurnGuard's cloud only ever receives usage metadata — tokens and cost — never your keys or prompt contents.
Anthropic, OpenAI, and Google(In View) today, unified into one dashboard with per-model breakdowns. More providers are on the way.
Effectively none — the proxy adds sub-millisecond overhead and streams responses straight through, so SSE and token streaming behave exactly as before.
Per request, from the provider's own token accounting and current model pricing — including cached and streamed tokens — so the number you see matches the invoice.
Stop checking your billing dashboard every morning.
Set a budget once. Let BurnGuard hold the line while you ship.
Protect Your Wallet