v0.1.1 · live

Every call.
Every model.
Logged.

Drop-in SDK that traces every LLM call across OpenAI, Anthropic, and Gemini. Token spend, latency, prompt versions — without a code change after install.

Read the docs ↗View on GitHub ↗
three lines

Drop in the SDK. That's it.

app/server.tscopy
// 1. install
import { tokentrace } from 'tokentrace-sdk';

// 2. init once at boot
tokentrace.init({ apiKey: process.env.TOKENTRACE_KEY });

// 3. every openai/anthropic/gemini call is now traced.
//    no other code change. zero overhead in dev.
const res = await openai.chat.completions.create({ ... });
features · 03

What the dashboard tells you.

spend
$2,418/24h
↓ 12.4%

Catch cost regressions before they ship.

Token spend tracked per model, per prompt version, per route. Anomalies surface within 60s.

latency
84ms p50
— stable

Find the prompt that just got slow.

Histogram + p50/p95/p99 per model. Drill into individual traces in two clicks.

prompts
142versions
↑ 3 today

Diff prompts the way you diff code.

Every prompt is versioned automatically. Side-by-side diff with cost and quality deltas.

pricing

Free under 100k traces. Flat fee above.

No per-seat tax. No “contact sales” tier.

Free100k traces / month
$0forever
  • ✓100k traces / month
  • ✓7 days retention
  • ✓All providers
  • ✓Anomaly alerts
  • ✓Community Discord
Get an API key
Teamup to 10M traces / month
$49/ month, flat
  • ✓10M traces / month
  • ✓90 days retention
  • ✓Prompt diff + version history
  • ✓Slack / PagerDuty alerts
  • ✓SSO + audit log
  • ✓Priority email
Start 14-day trial

Above 10M traces, $4 per additional 1M. No overage surprises — we email you at 80%.