# Langfuse

> Traces, evals and prompt management for LLM apps and agents

- Page on StartupList: https://startuplist.tech/launches/langfuse
- Website: https://langfuse.com/
- Category: LLM Developer Tools
- Price: Free + $29/mo
- Company: Langfuse (https://startuplist.tech/companies/langfuse)

## At a glance

- The problem it solves: Teams building AI agents and LLM applications cannot see what their models do in production or measure whether changes improve quality, cost and latency
- Who it's for: Businesses (B2B)
- How it finds customers: Launch sites, Affiliates and partners

## About

Langfuse is an open platform for tracing, evaluating and improving AI agents and LLM applications. It records every model call, tool invocation and retrieval step as a hierarchical trace, so teams can see what their AI did in production, track token usage, cost and latency, and use that production data to build evaluations and fixes.

The product covers the development loop. In production you trace and monitor an application with dashboards, LLM-as-judge evaluators, and alerts, and iterate on what monitoring shows. During development you build datasets, run experiments, and evaluate prompts, models and code variants before deploying. Prompts are managed separately from code, with versioning, deployments and rollbacks. A playground lets teams test prompts on real inputs and compare models side by side, and human annotation queues support review of traces and creation of golden datasets.

Instrumentation works through the Langfuse SDKs for Python and TypeScript, or through any OpenTelemetry exporter, which covers languages such as Go, Java, .NET, Ruby, PHP and Swift. The site lists more than 100 integrations with agent frameworks, model providers and gateways. Langfuse also offers an AI Gateway: teams can swap the base URL and API key to route OpenAI and Anthropic calls through Langfuse, with calls traced automatically and cost shown per user and key.

Every feature is available through a REST API, an SDK, a CLI and an MCP server, so coding agents can manage prompts, query traces and run evals. An in-app assistant called Halo investigates production data and takes actions after approval.

Langfuse is open source under the MIT license and can be self-hosted with Docker Compose, Kubernetes via Helm, or Terraform on AWS, GCP or Azure. It is also offered as a cloud service with data regions in the US, EU and Japan. The site states SOC 2 Type II and ISO 27001 reports and a HIPAA-ready region.

The site lists a free Hobby plan with 50,000 units a month and 30 days of data access, at no cost and with no credit card required. The Core plan is $29 a month with 100,000 units included and additional units at $8 per 100,000. The Pro plan is $199 a month, and the Enterprise plan is $2,499 a month with custom pricing available. Startups, education and open source projects can apply for discounts.

The Langfuse Academy offers free learning material on the AI engineering loop, including worked examples, and the docs are readable as Markdown for coding agents.

## Links

- [Langfuse on StartupList](https://startuplist.tech/launches/langfuse)
- [Website](https://langfuse.com/)