Langfuse

## Key capabilities

- **Tracing** — capture nested traces of LLM and agent runs with latency, token, and cost metadata.
- **Prompt management** — version and deploy prompts, with playground iteration.
- **Evaluation** — score runs with model-based and custom evaluators, plus datasets.
- **Open source** — self-hostable, with an OpenTelemetry-compatible ingestion path.

## How Langfuse compares

Langfuse competes with other LLM observability/eval tools in this directory:

| Tool | Type | Self-hostable | Notable for |
| --- | --- | --- | --- |
| **Langfuse** | Open-source observability + evals | Yes | Tracing, prompt management, and evals in one OSS platform |
| **Arize Phoenix** | Open-source observability + evals | Yes | OpenTelemetry-based, local-first |
| **LangSmith** | Proprietary observability + evals | Enterprise tier | Deep LangChain / LangGraph integration |
| **Helicone** | Observability via a request-logging proxy | Yes | One-line proxy integration |

Choose Langfuse for an all-in-one open-source platform (tracing + prompts + evals); Phoenix for OpenTelemetry-native local tracing, LangSmith for the LangChain ecosystem, or Helicone for proxy-based logging.

## Editorial notes

Langfuse is a strong option when teams want open-source LLM observability with self-hosting and prompt lifecycle support.

## Disclosure

Editorial listing. No paid placement or affiliate link is used.