llm-cost-meter — Per-feature, per-user cost attribution for LLM API calls

The Problem

Your LLM bill is a black box

Most teams have one API key, one invoice, and zero visibility into which features or users drive cost.

Without llm-cost-meter

One API key, one invoice, no breakdown
CFO asks "where is this $3,000 coming from?" — "We're not sure."
Can't optimize what you can't see

With llm-cost-meter

Cost per feature: chat $2.40, summarizer $0.31, classifier $0.02
Cost per user: Alice $1.80, Bob $0.62, Charlie $0.31
Insight: Chat drives 89% of cost but only 25% of calls

How It Works

Three steps. Two minutes.

Wrap your existing API calls. That's it. No proxies, no config servers, no infrastructure changes.

Wrap

const response = await meter(
  () => client.messages.create({
    model: 'claude-sonnet-4-20250514',
    max_tokens: 1024,
    messages: [{ role: 'user', content: '...' }]
  }),
  { feature: 'chat', userId: 'user_123' }
);

See

[llm-cost-meter] chat — $0.0105 (1,500 tokens, 820ms)

Report

Feature             Calls   Total Cost
chat                312     $3.78
article-summarizer  843     $3.29
tag-classifier      129     $0.02
TOTAL               1,284   $7.09

Features

Everything you need, nothing you don't

🔍

Auto-Detection

Detects OpenAI and Anthropic automatically from response shape

⚡

Streaming Support

meterStream() tracks cost from streaming responses

🛠

Express Middleware

Drop-in createExpressMiddleware() for your routes

📊

22+ Models

Built-in pricing for GPT-4o, Claude Opus, o1, o3, and more

💻

Interactive Dashboard

Charts, filters, drill-down, CSV export at localhost:3000

📄

CLI Reports

Group by feature, user, model. Export CSV, JSON.

🔔

Webhook Adapter

POST events to Slack, Zapier, or any URL. Single or batch mode.

📡

OpenTelemetry

Export cost metrics to Datadog, New Relic, Honeycomb, Grafana.

🚨

Budget Alerts

Daily cost limits per feature. Get notified when thresholds are exceeded.

Code Examples

Fits into your existing code

No SDK rewrites. No config files. Just wrap your calls.

import { meter, configure } from 'llm-cost-meter';
configure({ adapters: ['console', 'local'] });

const response = await meter(
  () => openai.chat.completions.create({
    model: 'gpt-4o',
    messages: [{ role: 'user', content: 'Hello' }]
  }),
  { feature: 'chat', userId: req.user.id }
);

import { meterStream } from 'llm-cost-meter';

const stream = await meterStream(
  () => openai.chat.completions.create({
    model: 'gpt-4o',
    messages: [{ role: 'user', content: 'Write a poem' }],
    stream: true
  }),
  { feature: 'chat', userId: 'user_123' }
);

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? '');
}
// Cost tracked automatically when stream ends

import { createExpressMiddleware } from 'llm-cost-meter';

app.post('/api/chat',
  createExpressMiddleware({ feature: 'chat' }),
  async (req, res) => {
    const response = await req.meter(() =>
      client.messages.create({ ... })
    );
    res.json(response);
  }
);

import { configureBudget, WebhookAdapter, OTelAdapter } from 'llm-cost-meter';

// Budget alerts — get notified when costs exceed limits
configureBudget({
  rules: [
    { feature: 'chat', dailyLimitUSD: 50,
      onExceed: (rule, spent) =>
        sendSlack(`Chat exceeded $${rule.dailyLimitUSD}/day! Spent: $${spent.toFixed(2)}`)
    },
    { feature: '*', dailyLimitUSD: 200,
      onExceed: (rule, spent) => pageOncall('Global LLM budget exceeded')
    },
  ]
});

// Webhook — POST events to Slack, Zapier, or any URL
configure({
  adapters: ['local', new WebhookAdapter({
    url: 'https://hooks.slack.com/services/...',
    batchSize: 10, flushIntervalMs: 5000
  })]
});

// OpenTelemetry — export to Datadog, New Relic, Honeycomb
configure({ adapters: ['local', new OTelAdapter()] });

Comparison

How it stacks up

No proxies. No gateways. No infrastructure to maintain.

Feature	API Keys per Service	AI Gateways	llm-cost-meter
Cost per feature	✗ No	✗ No	✓ Yes
Cost per user	✗ No	✗ No	✓ Yes
Cost per call	✗ No	✓ Yes	✓ Yes
Cross-provider	✗ No	● Partial	✓ Yes
No vendor lock-in	✓ Yes	✗ No	✓ Yes
Setup time	None	Hours	2 minutes
Budget alerts	✗ No	● Partial	✓ Yes
OpenTelemetry export	✗ No	● Partial	✓ Yes
Webhook notifications	✗ No	✗ No	✓ Yes
Price	Free	$$$	Free

Start tracking in 2 minutes

$ npm install llm-cost-meter

import { meter, configure } from 'llm-cost-meter';
configure({ adapters: ['console', 'local'] });

const response = await meter(
  () => openai.chat.completions.create({ model: 'gpt-4o', messages }),
  { feature: 'chat', userId: user.id }
);

GitHub npm Documentation Dashboard Demo

Know exactly where yourAI budget goes.