Langfuse

What it does

Langfuse is an open source platform for developing, monitoring, evaluating, and debugging LLM applications. The GitHub README describes tracing of LLM calls along with retrieval, embedding, and agent actions, with inspection of logs and user sessions through a web UI and demo project.

Core product areas listed in the README include LLM application observability, prompt management with versioning and caching, evaluations using LLM-as-a-judge, code evaluators, user feedback, and manual labeling, plus datasets for test sets and experiments. A playground links traced outputs to prompt iteration, and a comprehensive API with Python and JS/TS SDKs supports custom workflows.

Deployment options cover Langfuse Cloud with a free tier and self-hosting via Docker Compose, virtual machines, Kubernetes with Helm, and Terraform templates for AWS, Azure, and GCP. The homepage notes support for OpenTelemetry instrumentation and 100+ integrations across languages, frameworks, and model providers.

Verified capabilities

Quick start

Log your first LLM call with the Python SDK

Create a Langfuse project and API credentials, install the SDK, set project keys, and trace a function with the observe decorator and OpenAI integration as shown in the README quickstart.

pip install langfuse openai

from langfuse import observe
from langfuse.openai import openai

@observe()
def story():
    return openai.chat.completions.create(
        model="gpt-4o",
        messages=[{"role": "user", "content": "What is Langfuse?"}],
    ).choices[0].message.content

Source: https://github.com/langfuse/langfuse. Examples use placeholders only. Never paste a real key into a profile, config file you share, or a ticket.

MCP support: The Langfuse homepage states that SKILL.md, CLI, and MCP connect coding agents to Langfuse for work in the app or IDE.

Works with

Only sourced support is listed. A missing framework means AgentsUse has not verified it yet, not that it cannot work.

Health and maintenance

Pricing and license

Freemium model with a free Hobby tier for Langfuse Cloud and paid Core, Pro, and Enterprise tiers. Cloud pricing is based on ingested units defined as traces plus observations plus scores, with graduated per-unit rates. Self-hosted open source use is free under MIT, while self-hosted Enterprise features are commercially licensed.

Official pricing or docs →

Limitations and safety

• Self-hosting requires operating Docker or Kubernetes infrastructure including Postgres, ClickHouse, Redis, and object storage components described in the self-hosting docs.
• Enterprise security and platform features such as SCIM, audit logs, and data retention policies remain under a commercial license, separate from the MIT open source core.
• Cloud cost scales with total units, so a single agent run with many observations and scores counts as multiple billable units.

Alternatives to Langfuse

Related tools

Common questions