Langtrace
Open-source, OpenTelemetry-based observability for LLM apps — traces, evals, metrics.
About
Langtrace is an open-source observability platform for LLM applications built on OpenTelemetry standards. One line of SDK code traces LLM API calls, retrievals, vector-database operations, and framework executions — capturing prompts, responses, token counts, cost, and latency — with evaluations and analytics on top. It ships TypeScript and Python SDKs with first-class auto-instrumentation for frameworks like DSPy.
Built by Scale3 Labs, Langtrace is available as a managed cloud (app.langtrace.ai) or fully self-hosted via Docker for teams that need data to stay on their own infrastructure. It sits in the same category as Langfuse, Helicone, and Arize Phoenix, differentiating on OTEL-native ingestion and a permissive open-source model. For teams iterating on LLM features, it pairs tracing with prompt registries and cost analytics rather than just raw spans.
What it can do
- Auto-instrumented tracing for LLM calls and RAG pipelines
- Token, cost, and latency analytics
- Evaluations and prompt registry
- Self-hosted Docker deployment or managed cloud
Setup
Install the SDK (npm i @langtrase/typescript-sdk or pip install langtrace-python-sdk) and init with an API key.
Prerequisites
An LLM application in TypeScript or Python; an API key from langtrace.ai for cloud use.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 60; Trending score 68/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.