Save 80% of integration time
LLM Observability Without the Lock-in
Capture LLM traces once and route them to any observability platform. End vendor lock-in, reduce costs, and simplify your AI infrastructure with intelligent routing.
Get started in seconds
Works with OpenAI, Anthropic, LangChain, and 30+ providers
pip install untrace-sdk
Instrument once, observe everywhere
Zero-latency observability that just works
// Without Untrace - flying blind 🙈
const response = await openai.chat.completions.create({
model: 'gpt-4',
messages: [{ role: 'user', content: prompt }],
});
// ❌ No visibility into:
// - Only one LLM Observability tool
// - Token usage & costs
// - Response latency
// - Error rates
// - Model performance// With Untrace - full observability 🔍
import { init } from '@untrace/sdk';
// One line setup! // [!code highlight]
init({ apiKey: 'your-api-key' }); // [!code highlight]
// ✅ Automatic tracking of: // [!code highlight]
// - Token usage & costs ($0.03) // [!code highlight]
// - Response latency (1.2s) // [!code highlight]
// - Success/error rates // [!code highlight]
// - Complete traces // [!code highlight]
const response = await openai.chat.completions.create({
model: 'gpt-4',
messages: [{ role: 'user', content: prompt }],
});
Intelligent LLM Trace Routing
Capture traces once, route them intelligently to multiple observability platforms based on your rules.
Universal Trace Capture
Capture LLM traces through OpenAI-compatible proxy, native SDKs, or webhooks. Support for streaming and non-blocking capture.
10+ Platform Integrations
Pre-built integrations for LangSmith, Langfuse, Keywords.ai, Helicone, and more. Custom webhook support for any platform.
Smart Routing Rules
Route traces based on model type, cost thresholds, errors, or custom metadata. Sample intelligently to reduce costs.
Real-time Monitoring
Monitor trace flow, destination health, and costs in real-time. Debug failed deliveries and test new routes.
Enterprise-Grade Infrastructure
Built for scale, security, and reliability. Handle millions of traces with sub-50ms latency.
SOC2 Prepared
Enterprise-grade security with encryption in transit and at rest. Automatic PII detection and redaction.
Global Edge Network
Multi-region deployment with automatic failover. Sub-50ms P95 latency overhead globally.
Frequently Asked Questions
Everything you need to know about Untrace. Can't find what you're looking for? Contact our support team.