Langfuse

Platform

Open-source tracing, evals and prompt management for LLM apps

Price: Free tier, then paid
Access: Public and secret API keys

About

Open-source LLM engineering platform for tracing model calls, running evaluations and versioning prompts, built on OpenTelemetry. Use Langfuse Cloud in the US, EU or Japan, or self-host the MIT-licensed core; some enterprise features need a paid license.

What you can do with it

Trace every LLM call, tool call and agent step with cost and latency
Score outputs with LLM-as-a-judge, datasets and human annotation queues
Version prompts and fetch them at runtime without redeploying

Get started

Create a Langfuse Cloud project or self-host Langfuse
Create API keys in the project settings

Example

# pip install langfuse openai
# export LANGFUSE_PUBLIC_KEY, LANGFUSE_SECRET_KEY, LANGFUSE_BASE_URL, OPENAI_API_KEY
from langfuse.openai import openai

completion = openai.chat.completions.create(
  name="test-chat",
  model="gpt-4o",
  messages=[{"role": "user", "content": "1 + 1 = "}],
)

Details

Hosting: Hosted service, Self-hosted
Available in: Worldwide
Official SDKs: Python, JavaScript/TypeScript
MCP server: Remote