Langfuse is an open-source LLM observability and evaluation platform, now part of ClickHouse. See 2026 pricing, tracing and eval features, and alternatives.
Category
AI Infrastructure & MLOps
Pricing
Freemium, from Free (Hobby, self-hosted or cloud); paid plans from $29/month
Langfuse is an open-source LLM engineering platform for tracing, evaluating, and managing prompts in production AI applications, built by a Berlin-based founding team that went through Y Combinator's W23 batch.
In January 2026, Langfuse was acquired by the analytical database company ClickHouse, whose infrastructure already powered Langfuse's own architecture; the platform remains MIT-licensed and open source for its core features, with Langfuse Cloud continuing to operate as a standalone service.
Key Features
Langfuse's core tracing engine captures every LLM call, retrieval step, tool execution, and custom logic within a request, along with timing, inputs, outputs, and metadata, giving engineers full visibility into application behavior.
The platform also includes prompt management with versioning, an evaluation system supporting human feedback and LLM-as-a-judge scoring, datasets and experiments for comparing prompt or model versions, and a playground linked directly to production traces.
Pricing
Langfuse offers a free Hobby tier with 50,000 units per month and 30-day retention, a Core tier at $29/month with 100,000 units and 90-day retention, a Pro tier at $199/month with three years of retention and compliance reports, and an Enterprise tier at $2,499/month with audit logs, SCIM, and dedicated support.
Usage beyond each tier's included allowance is billed on a graduated per-100,000-unit rate, and the fully open-source core can also be self-hosted for free under the MIT license.
Key Features
Distributed LLM tracing — Captures every LLM call, retrieval step, tool execution, and custom logic within a request, with full inputs, outputs, and timing.
Prompt management — Centrally version, collaboratively edit, and iterate on prompts outside of application code.
Evaluations and LLM-as-a-judge — Supports human feedback, user signals, and automated LLM-as-a-judge scoring for measuring output quality.
Datasets and experiments — Test and compare prompt or model versions against curated datasets before shipping changes to production.
Interactive playground — Test prompts directly and link playground sessions back to real production traces for context.
Framework-agnostic integrations — Integrates with the OpenAI SDK, LangChain, LlamaIndex, LiteLLM, and the OpenTelemetry ecosystem.
Self-hosted or managed cloud — Runs as a fully managed cloud service or a self-hosted Docker deployment under the MIT-licensed open-source core.
Enterprise compliance and controls — Higher tiers add SOC2/ISO27001 reports, audit logs, SCIM, SSO, and custom rate limits and SLAs.
Pros & Cons
Pros
Open-source, MIT-licensed core that can be self-hosted for free with production-scale usage
Framework-agnostic integrations mean it works without requiring LangChain specifically
Comprehensive feature set covering tracing, prompt management, evaluation, and datasets in one platform
Backed by ClickHouse's infrastructure and resources following the January 2026 acquisition
Trusted at large scale, with usage across a significant share of Fortune 500 companies
Cons
Enterprise-grade features like audit logs and SCIM require the highest, $2,499/month tier
Usage-based overage pricing can be harder to predict than flat-rate competitors for high-volume teams
Recent acquisition by ClickHouse may raise questions about long-term roadmap independence for some users
Self-hosting the full platform still requires infrastructure and DevOps effort to maintain
Free Hobby tier is limited to two users and 30 days of data retention
Pricing
Hobby Free Monthly
Core $29/month Monthly
Pro $199/month Monthly
Enterprise $2,499/month Yearly commitment
Frequently Asked Questions
Is Langfuse free to use?
Yes, Langfuse offers a free Hobby cloud tier and a fully open-source, self-hostable core under the MIT license. Paid Core, Pro, and Enterprise plans add higher usage limits, longer retention, and enterprise features.
Who owns Langfuse?
Langfuse was founded in 2022 by Maximilian Deichmann, Marc Klingen, and Clemens Rawert, and was acquired by the analytical database company ClickHouse in January 2026.
Does Langfuse require LangChain to work?
No, Langfuse is framework-agnostic and integrates with the OpenAI SDK, LangChain, LlamaIndex, LiteLLM, and OpenTelemetry, so it does not require using LangChain specifically.
Can Langfuse be self-hosted?
Yes, Langfuse's core tracing, prompt management, API, and data model are MIT-licensed and can be self-hosted via Docker at production scale for free.
What does Langfuse cost?
Langfuse's cloud pricing ranges from a free Hobby tier to $29/month Core, $199/month Pro, and $2,499/month Enterprise, with usage-based overage billing beyond each tier's included allowance.
What are the main alternatives to Langfuse?
Common alternatives include LangSmith, Helicone, Arize Phoenix, Weights and Biases Weave, and Braintrust.