Langfuse
Free planOpen-source LLM observability platform
An open-source LLM observability platform offering tracing, evaluation, prompt management, and cost tracking for LLM applications. It integrates with OpenTelemetry, LangChain, the OpenAI SDK, LiteLLM, and more to monitor production AI apps.
vs. similar tools: Its strength is the ability to self-host the MIT-licensed core, running tracing, evaluation, and prompt management without your data ever leaving your environment.
Overview
At a glance
- Tracing, evaluation, prompt management, and cost tracking in one platform
- Self-host the MIT core so data never leaves your environment
- Broad integrations with OpenTelemetry, LangChain, OpenAI SDK, and more
- Weak Korean support, so guidance and UI lean English
- Production traffic incurs extra cost once units run over
- Best for: Teams observing production LLM apps on their own infrastructure
Read more
Langfuse is an open-source observability platform that provides tracing, evaluation, prompt management, and cost tracking for LLM applications in one place. Integrating with OpenTelemetry, LangChain, the OpenAI SDK, LiteLLM, and more, it fits development teams and MLOps owners who want to see how their production AI apps actually behave.
Its strength lies in depth and operational freedom. With an observability-depth rating of 90, it can trace and evaluate down to individual calls, and an ecosystem rating of 88 means it slots into an existing stack without friction. Above all, you can self-host the MIT-licensed core, and the biggest differentiator is running it in your own environment without sending sensitive trace data outside. An API is also offered, so evaluation and prompt management can be woven into your own workflow.
That said, Korean support is on the weaker side, so docs and screens are English-centric. If you use the cloud, production traffic beyond the free tier adds cost per unit consumed, so it helps to gauge your traffic volume in advance.
Those who specifically need a Korean UI, or who want a finished SaaS usable immediately without code integration, are better off considering other tools as well.
Pricing
| Plan | Monthly price | Limits |
|---|---|---|
| Hobby | $0/mo | 50,000 units free per month; self-hosting is free under the MIT license |
| Core | $29/mo | For production traffic, with overage units at 8 USD per 100,000 |
| Pro | $199/mo | Long-term retention, SOC2/ISO27001 reports, and higher rate limits |
Specs
- API
- Yes
- Open source
- Yes
- Self-hosting
- Available
- Korean support
- Not supported
- Commercial use
- Allowed
Popularity
Buzz and recognition on absolute thresholds
Absolute-threshold score
77
High confidence6/6 signals
Each axis maps to a 1-10 absolute threshold where 10 means broadly recognizable. Collected: 2026-06-13.
Verified public benchmark: 2,300+ customers and 10B+ monthly observations reported by Langfuse (as of 2026-06-13) Source
Related tools
By popularity
- Hugging Face
Its strength is being the de facto standard hub, offering hundreds of thousands of public models and datasets alongside inference and deployment infrastructure on a single platform.
- OpenRouter
Its strength is passing through provider pricing as-is while offering automatic fallback and model routing from a single key.
- AssemblyAI
It goes beyond transcription by packaging natural-language prompting, keyterm boosts, medical mode, and voice agent APIs in one platform.
- Deepgram
It focuses on realtime voice-agent infrastructure, including turn detection and interruption handling, beyond STT and TTS.
- Glean
It goes beyond document search by packaging permissions, connectors, agents, and developer integrations for enterprise-wide rollout.
Compare Langfuse
Last updated: 2026-05-30
All tools