Install & onboarding

• A GitHub account (we use GitHub OAuth for sign-in)
• A Node.js project, or anything that can make an HTTP POST (Python, Go, curl)
• Your OpenAI / Anthropic / Bedrock / whatever LLM provider keys (stay on your side — we never see them)
• 5 minutes

Click below. We use GitHub OAuth — no password to remember. A workspace is created automatically on first sign-in.

Right after sign-in you land on the guided onboarding screen. Name your key (the default is fine), click Create key, and copy it once — the full key is shown one time only. Store it somewhere you can find it (1Password, your .env file).

llm-cost-meter is the open-source SDK that wraps your LLM calls and ships cost events to your workspace. Node / Next.js / Bun all work.

npm install llm-cost-meter@^0.6.0

Last verified version: v0.6.0. Source on GitHub. Python and Go SDKs are roadmap — until they land, use the curl path in step 4.

The onboarding screen is polling — your dashboard flips green the moment the first event lands. Pick whichever path matches your stack.

Node — wrap an LLM call with track()

import { configure, track } from "llm-cost-meter";
import OpenAI from "openai";

configure({
  adapters: ["local", "cloud"],
  cloudApiKey: process.env.TRAICE_API_KEY,
});

const openai = new OpenAI();

// Wrap the call. tenantId is the wedge — pass the customer id you bill.
await track(
  { feature: "chat", tenantId: customer.id },
  () => openai.chat.completions.create({
    model: "gpt-4o-mini",
    messages: [{ role: "user", content: "hello" }],
  }),
);

Anything else — POST directly

Works from Python (requests), Go, Ruby, shell — anything that can send a JSON body.

curl -X POST "https://ll-meter-cloud.vercel.app/api/v1/events" \
  -H "authorization: Bearer YOUR_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "provider": "openai",
    "model": "gpt-4o-mini",
    "feature": "chat",
    "tenantId": "customer_42",
    "promptTokens": 100,
    "outputTokens": 50,
    "costUsd": 0.0012
  }'

The dashboard fills in as events arrive. To unlock the headline number (AI gross margin per customer), you also need to tell us what each customer pays you:

1. Send events with a tenantId per call (step 4).
2. Go to Settings → Customer revenue. Paste a row per tenantId with their monthly revenue in USD. (Manual paste for beta — Stripe import is on the roadmap once we settle the payments vendor question.)
3. Refresh /dashboard. The “Top AI-margin leaks” panel and per-tenant table now show who's profitable on AI and who's a subsidy in disguise.

Want a preview of what this looks like with real numbers? Open the live demo →

• Free tier:1,000 events / month. You'll see a banner on the dashboard at 80% and again if you hit 100% (after which the ingest API returns HTTP 402 until next month).
• Auth: GitHub OAuth only during beta. Email / password / Google are Tier-1, after we have signed paying customers.
• Upgrades: No self-serve billing yet — Stripe is blocked on the Israeli-merchant question. If you want to pay, email usand we'll flip your workspace plan manually. Free during beta unless you ask otherwise.
• Domain: We're at ll-meter-cloud.vercel.app while we cut over to runtraice.com.
• Out of scope on purpose: in-request gateway, prompt playground, distributed tracing. See what trAIce doesn't do →

Beta is a two-way deal. You get the product and direct access; we get to watch how you actually use it. If anything is unclear or broken:

• Email shmulikdavar@gmail.com — same-day reply during beta.
• Found a bug in the SDK? File it on the GitHub repo.