Capability · Framework — observability

Langtrace

Langtrace (by Scale3 Labs) is an open-source LLM observability system that speaks OpenTelemetry end-to-end. Its SDK captures spans for LLM calls, vector DBs, and frameworks like LangChain, LlamaIndex, and CrewAI, and sends them to either Langtrace Cloud or any OTel collector. Differentiators include built-in evals, prompt playground, and Anthropic computer-use support.

Framework facts

Category: observability
Language: Python / TypeScript
License: AGPL-3.0
Repository: https://github.com/Scale3-Labs/langtrace

Install

pip install langtrace-python-sdk
# Self-host the UI
git clone https://github.com/Scale3-Labs/langtrace && cd langtrace && docker compose up

Quickstart

from langtrace_python_sdk import langtrace
langtrace.init(api_key='LANGTRACE_KEY')

from openai import OpenAI
OpenAI().chat.completions.create(model='gpt-4o', messages=[{'role':'user','content':'hi'}])

Alternatives

Langfuse — OSS
Arize Phoenix — OSS
OpenLLMetry — OTel SDK

Frequently asked questions

Langtrace or Langfuse?

Both are OSS and solid. Langtrace is OTel-native (any OTel backend works), Langfuse has a slightly richer evals UI. If you're already on OpenTelemetry, Langtrace slots in more cleanly.

Is AGPL a problem?

Only if you host Langtrace as a public SaaS on top of it; normal in-company or self-hosted use is fine. Enterprise license is available.

Sources

Langtrace docs — accessed 2026-04-20
Langtrace GitHub — accessed 2026-04-20