Langfuse is an open-source platform for tracing, evaluating, and improving LLM applications and agents.
Core capabilities
Langfuse records traces, sessions, users, token use, and cost. The pricing comparison lists SDKs for Python and JavaScript, OpenTelemetry, prompt versioning, a playground, experiments, and human annotation. The homepage describes using production data to see how an agent behaved and to fix quality, cost, and latency. You can use Langfuse Cloud or self-host. Self-hosting docs linked from the pricing page include Docker Compose, Kubernetes, and Terraform.
How to use it
Sign up for the cloud app or run a self-hosted instance. Send traces from the SDK or OpenTelemetry, then inspect a trace, version a prompt, or run an experiment on saved inputs. Hobby can be started without a credit card.
Who it suits
It suits developers who already call a model and need observability and evaluation. It does not itself replace the model API you call.
Pricing, access and limits
Cloud Hobby is free, with 50,000 units a month, 30 days of data access, 2 users, and all platform features within those limits. Core is $29 per month, with 100,000 units included, 90 days of data access, and unlimited users. Pro is $199 per month, with 3 years of data access, higher rate limits, and SOC 2 and ISO 27001 reports. Extra units on Core and Pro are listed at $8 per 100,000 units. A Teams add-on is $300 per month. Enterprise is $2,499 per month. The page also offers self-hosting.
Frequently asked questions
Is Langfuse only a log viewer?
The site also covers prompt management, experiments, and human review of traces, in addition to tracing.
Can I run it myself?
Yes. The pricing page has a self-hosted option beside Langfuse Cloud.
