Website · Docs · Dashboard · Twitter

🎬 See it in action

https://github.com/user-attachments/assets/56962fc8-b9cf-49b2-9481-bc10aca6fb56

The Problem

You're running OpenAI, Anthropic, Stripe, Supabase, SendGrid, and a dozen other APIs. Your monthly bill is $2,847 and climbing 340% month-over-month.

You have no idea which feature is burning money.

Observability platforms charge $199/mo. API monitoring tools charge $149/mo. Cost management SaaS charges $99/mo.

burn0 is free. One import. That's it.

burn0 ▸ $4.32 today (47 calls) ── openai: $3.80 · anthropic: $0.52

Quick Start

npm i @burn0/burn0

Add one line to your entry file:

import "@burn0/burn0"; // Must be first import

import express from "express";
import OpenAI from "openai";
// ... your app runs exactly the same

That's it. Costs appear in your terminal:

burn0 ▸ $0.47 today (12 calls) ── openai: $0.41 · stripe: $0.06

On exit:

burn0 ▸ session: $0.47 (12 calls, 4m 22s) ── today: $14.32 ── ~$430/mo

Want full history and dashboards? Add an API key:

# 1. Sign in with GitHub and create a key
open https://burn0.dev/login
#    → Dashboard → API Keys → Create Key

# 2. Add it to your project
echo 'BURN0_API_KEY=b0_sk_your_key_here' >> .env

# 3. Restart — costs now sync to burn0.dev

Now you get a live event feed, cost breakdown by service, monthly projections, and full request history — all at [burn0.dev/dashboard](https://burn0## The Problem.dev/dashboard).

burn0 only syncs metadata (service, model, tokens, cost, latency) — never request/response bodies or your API keys.

How It Compares

| | Observability Platform | API Monitoring | Cost Management SaaS | burn0 | | ------------------- | ---------------------- | -------------- | -------------------- | ---------------------- | | Price | $199/mo | $149/mo | $99/mo | Free forever | | Setup | SDK + dashboard config | Proxy setup | Manual tagging | One import | | Latency | 5-50ms | 10-100ms | Async | <1ms | | Per-feature | Manual instrumentation | No | Manual | burn0.track() | | Works locally | No | No | No | Yes | | Open source | No | No | No | MIT Licensed | | Data leaves app | Always | Always | Always | Only if you opt in |

You're spending $526/mo on tools that burn0 replaces for $0.

CLI

# Interactive setup wizard
npx burn0 init

# Cost report (last 7 days)
npx burn0 report

# Today only
npx burn0 report --today

# Run any app with tracking (zero code changes)
npx burn0 dev -- node app.js

# Connect to cloud dashboard
npx burn0 connect

burn0 report output

burn0 report ── last 7 days

Total: $12.47 (342 calls)

openai         $8.32   ██████████████░░░░░░  67%
anthropic      $3.15   ██████░░░░░░░░░░░░░░  25%
google-gemini  $0.85   ██░░░░░░░░░░░░░░░░░░   7%
resend         $0.15   ░░░░░░░░░░░░░░░░░░░░   1%

── projection ─────────────────────────────
~$53/mo estimated (based on last 7 days)

Feature Attribution

Know exactly which feature burns money:

import { track } from "@burn0/burn0";

await track("onboarding", async () => {
  const profile = await ai.generateProfile(user);
  await stripe.createSubscription(user.id);
  await sendWelcomeEmail(user.email);
});

burn0 ▸ feature "onboarding" ── $0.47/user
         └─ openai     $0.39  (83%)
         └─ stripe     $0.0001 (2%)
         └─ sendgrid   $0.08  (17%)

Track per user, per feature, per request. No manual tagging. No dashboards to configure.

50+ Services Supported

burn0 auto-detects services from hostnames. Zero configuration.

AI / LLMs

| Service | Detection | Pricing Model | | ------------- | ----------------------------------- | ----------------- | | OpenAI | api.openai.com | Per-token (exact) | | Anthropic | api.anthropic.com | Per-token (exact) | | Google Gemini | generativelanguage.googleapis.com | Per-token (exact) | | Mistral | api.mistral.ai | Per-token (exact) | | Cohere | api.cohere.ai | Per-token | | Groq | api.groq.com | Per-token | | Together AI | api.together.xyz | Per-token | | Perplexity | api.perplexity.ai | Per-token | | DeepSeek | api.deepseek.com | Per-token | | Replicate | api.replicate.com | Per-second | | Fireworks AI | api.fireworks.ai | Per-token | | AI21 Labs | api.ai21.com | Per-token | | Pinecone | *.pinecone.io | Per-request |

Pay-per-use APIs

| Service | Detection | Pricing Model | | ----------- | --------------------- | --------------- | | Stripe | api.stripe.com | Per-transaction | | PayPal | api.paypal.com | Per-transaction | | Plaid | *.plaid.com | Per-request | | SendGrid | api.sendgrid.com | Per-email | | Resend | api.resend.com | Per-email | | Twilio | api.twilio.com | Per-message | | Vonage | api.nexmo.com | Per-message | | Algolia | *.algolia.net | Per-search | | Google Maps | maps.googleapis.com | Per-request | | Mapbox | api.mapbox.com | Per-request | | Cloudinary | api.cloudinary.com | Per-transform | | Sentry | sentry.io | Per-event | | Segment | api.segment.io | Per-event | | Mixpanel | api.mixpanel.com | Per-event |

Databases & Infrastructure

| Service | Detection | Pricing Model | | ------------- | ------------------------ | -------------- | | Supabase | *.supabase.co | Per-request | | PlanetScale | *.psdb.cloud | Per-request | | MongoDB Atlas | *.mongodb.net | Per-request | | Upstash | *.upstash.io | Per-request | | Neon | *.neon.tech | Per-request | | Turso | *.turso.io | Per-request | | Firebase | *.firebaseio.com | Per-request | | AWS S3 | *.s3.amazonaws.com | Per-request | | AWS Lambda | lambda.*.amazonaws.com | Per-invocation | | Vercel | api.vercel.com | Per-request |

Unknown APIs are auto-tracked by request count. Nothing slips through.

How It Works

Your app starts
  │
  ├─ import '@burn0/burn0' patches globalThis.fetch + node:http
  │
  ├─ Every outbound HTTP call is intercepted (zero behavior change)
  │
  ├─ Service identified from hostname (api.openai.com → OpenAI)
  │
  ├─ Token counts + costs extracted from response metadata
  │
  └─ Costs displayed in terminal + stored in local ledger

1. Interception is synchronous — your request goes out immediately
2. Cost extraction is async — happens after the response, never blocks
3. Sub-millisecond overhead — benchmarked, not estimated
4. Never reads content — only extracts metadata: service, model, tokens, status, latency
5. Never throws — graceful degradation if anything fails internally
6. ±2% accuracy — exact token counts from LLM APIs, bundled pricing for SaaS

Two Modes

| Mode | API Key | What happens | | ------------------- | ------- | ------------------------------------------------------------------------------------------------------------- | | Local (default) | No | Costs in terminal + local ledger. Zero network calls to burn0. | | Cloud (opt-in) | Yes | Same as local + events sync to dashboard for full history and team visibility. |

Start local, upgrade when ready. burn0 works perfectly without an API key. When you want history and dashboards, add a key — it takes 60 seconds. Get your free API key →

Configuration

| Env Variable | Default | Description | | ------------------- | ----------------------- | --------------------------------------- | | BURN0_API_KEY | — | API key for cloud mode | | BURN0_API_URL | https://api.burn0.dev | Backend URL | | BURN0_DEBUG | false | Enable debug logging | | BURN0_ENABLE_TEST | — | Set to 1 to enable in NODE_ENV=test |

Works With Everything

burn0 works with any Node.js framework. If it makes HTTP calls, burn0 tracks the costs.

Express · Next.js · Fastify · Hono · Koa · NestJS · Remix · Nuxt

Requirements: Node.js >= 18

Frequently Asked Questions

Does it slow down my API calls?

No. Interception is synchronous but event processing is fully async. burn0 adds sub-millisecond overhead to your API calls.

Does it send my data anywhere?

By default, no. In local mode, costs are logged to your terminal and stored in a local file. Cloud mode (opt-in) ships only metadata — never request/response bodies.

How accurate are the cost estimates?

burn0 extracts exact token counts from LLM API responses. For pay-per-use APIs, it uses bundled pricing data. Accuracy is within ±2%.

Can I use it in production?

Yes. burn0 is designed for production use. It never throws, never adds latency, and gracefully degrades if anything fails internally.

Is it really free?

Yes. burn0 is MIT licensed and free forever. No API key required for local mode. Cloud features (dashboard, team analytics) are available as a paid tier.

Development

git clone https://github.com/burn0-dev/burn0.git
cd burn0
npm install
npm run build
npm test

Community

| Channel | Link | | ---------- | ----------------------------------------------------- | | 🌐 Website | burn0.dev | | 📖 Docs | docs.burn0.dev | | 🐦 Twitter | @burn0dev | | 💻 GitHub | burn0-dev/burn0 |

MIT License · Built by the burn0 team

⭐ If burn0 saves you money, consider starring the repo.