Langtrace

Langtrace is an open-source, OpenTelemetry-based observability tool for LLM apps, offering real-time tracing, cost and latency metrics, and evaluations across models, frameworks, and vector databases. Its standards-based traces export anywhere, making it ideal for teams avoiding vendor lock-in.

Langtrace is an end-to-end observability tool built on OpenTelemetry for monitoring LLM applications. Its SDK is a lightweight library you install and import to collect traces across popular LLMs, LLM frameworks, and vector databases, and because those traces follow the OpenTelemetry standard, you can export them to Langtrace or to any other observability stack without being locked into a proprietary format or even needing a Langtrace API key. The tool surfaces real-time monitoring of key metrics — costs and token usage, accuracy, and response times — along with detailed traces and logs across your stack. It also includes evaluation capabilities, letting developers manually score the outputs of an LLM application so they can measure and improve quality over time. It supports both Python and TypeScript, fitting the two most common languages for building LLM apps. Being open source and standards-based is Langtrace's core appeal: teams that prize portability and avoiding vendor lock-in can self-host and route their telemetry however they like, and it sits alongside tools like Langfuse and Arize Phoenix in the LLM observability space. The trade-off is that, as a leaner open-source project, it has a smaller ecosystem and fewer turnkey enterprise features than some heavily funded competitors.