Quick Start Guide
Three steps to complete AI cost visibility
Install the SDK
npm install @tokentra/sdk
# or
pip install tokentraInitialize TokenTra
import { TokenTra } from '@tokentra/sdk';
const tokentra = new TokenTra({
apiKey: 'tt_live_xxx'
});Wrap Your AI Client
import OpenAI from 'openai';
const openai = tokentra.wrap(new OpenAI());
// Use as normal - costs are tracked automatically
const response = await openai.chat.completions.create({
model: 'gpt-4',
messages: [{ role: 'user', content: 'Hello!' }]
}, {
tokentra: {
feature: 'chat',
team: 'product',
userId: 'user_123'
}
});That's it!
Your AI costs are now being tracked. Head to your TokenTra dashboard to see real-time spending data, set up alerts, and discover optimization opportunities.
SDK Features
Zero Latency Impact
All telemetry is sent asynchronously after the AI response is returned. Your API calls complete at exactly the same speed.
Non-Blocking
Batch telemetry collection ensures your application performance is never impacted by TokenTra.
Privacy First
We never see your prompts or responses. Only usage metrics and costs are collected.
Cost Attribution
Tag every AI request with custom attributes for granular cost tracking
// Add attribution to any request
const response = await openai.chat.completions.create({
model: 'gpt-4',
messages: [...]
}, {
tokentra: {
// Required: Feature being used
feature: 'customer-support-chat',
// Optional: Team responsible
team: 'support-engineering',
// Optional: Project or product
project: 'helpdesk-v2',
// Optional: End user (for per-user economics)
userId: 'user_abc123',
// Optional: Custom tags for your use case
metadata: {
environment: 'production',
region: 'us-west',
experimentId: 'exp_456'
}
}
});Track by Team
See which teams are spending the most and enable internal chargebacks.
Track by Feature
Understand the unit economics of every AI-powered feature in your product.
Track by User
Calculate cost per user and identify heavy users driving your AI spend.
Custom Metadata
Add any custom tags for A/B tests, environments, or business-specific tracking.
Supported Providers
Works with all major AI providers out of the box
OpenAI
GPT-5, GPT-5.1, GPT-5 Pro, o3, o3-mini, DALL-E 4, Whisper v3, Embeddings
Anthropic
Claude Opus 4.5, Claude Sonnet 4, Claude Haiku 4
Google Vertex AI
Gemini 3 Pro, Gemini 3 Flash, Gemini 3 Ultra, Imagen 3
Azure OpenAI
All OpenAI models via Azure deployments (GPT-5, o3)
AWS Bedrock
Claude Opus 4.5, Titan v2, Llama 3, Mistral Large 2, Stable Diffusion 3
Python SDK
Full feature parity with the Node.js SDK
from tokentra import TokenTra
from openai import OpenAI
# Initialize TokenTra
tokentra = TokenTra(api_key='tt_live_xxx')
# Wrap your OpenAI client
openai = tokentra.wrap(OpenAI())
# Use as normal
response = openai.chat.completions.create(
model='gpt-4',
messages=[{'role': 'user', 'content': 'Hello!'}],
tokentra={
'feature': 'chat',
'team': 'product',
'user_id': 'user_123'
}
)
# Async support
import asyncio
from openai import AsyncOpenAI
async_openai = tokentra.wrap(AsyncOpenAI())
async def main():
response = await async_openai.chat.completions.create(
model='gpt-4',
messages=[{'role': 'user', 'content': 'Hello!'}]
)
asyncio.run(main())