Why do customers choose Langfuse?

Customers choose Langfuse because we are...

• The most used open-source AI Engineering platform since 2024
• Part of ClickHouse, whose open-source analytics database has been used in production for over a decade. This makes Langfuse scalable, reliable, and cost-effective for production workloads, with strong financial backing
• Incrementally adoptable, start with one feature and expand to the full platform over time
• Model and framework agnostic, with 100+ integrations and OpenTelemetry support
• API-first, with skills for your coding agents and a UI people love using
• Serious about security and data control, with easy self-hosting on your own infrastructure
• Built with our community, with a transparent roadmap and open discussions on GitHub

See more of the companies building with Langfuse.

Full-stack open-source

Langfuse is open source, including the UI and backend.

• Self-host the same scalable backend that powers Langfuse Cloud. No Enterprise license needed to scale.
• All product capabilities—tracing, evaluations, prompt management, experiments, annotation, the playground, and more—are MIT licensed without any usage limits.
• We are transparent about what's on the roadmap and how we operate.
• We iterate with the community on GitHub because best practices for building great LLM applications are rapidly evolving. Please share your feedback and ideas with us.
• Keep data on your infrastructure or let Cloud handle operations. You can move between OSS, Enterprise self-hosting, and Cloud as your needs change.

Built for scale on ClickHouse

Observability is a data-intensive task. We run on ClickHouse, which is becoming a standard for observability. Anthropic and OpenAI rely on it for their large-scale workloads.

For Langfuse, this brings four benefits:

• Scalable: Langfuse Cloud already stores petabytes of tracing data.
• Fast: By optimizing Langfuse around ClickHouse, we increased performance by up to 165× for some queries (technical blog).
• Cost-effective: We are significantly cheaper than other solutions. Learn more.
• Reliable: ClickHouse is open source and has been used in production for over 10 years.

Built for complex agents

Agents take detours. We help you follow them.

• Follow nested model calls, tools (including MCP), and sub-agents through agent graphs and a zoomable timeline.
• Keep conversations together with sessions, and add your app's context with metadata and tags.
• Inspect images, audio, video, and documents in multimodal traces. Evaluate them with a compatible judge model.

See Langfuse in action for chat agents, coding agents, and workflow automation.

A complete platform. Start where you need us.

The pieces work together. You turn a production trace into a test case, compare prompt versions, and measure the results after you ship.

• Observability: understand what happened in each run.
• Prompt management: version, test, deploy, and roll back prompts.
• Evaluation: run experiments on datasets and score production outputs with human review, LLM judges, or code.
• Metrics and data: dashboards, alerts, and exports for quality, cost, and latency.

At Slite, connecting tracing and evaluations helped move experiments from quarterly to weekly.

We share how to put it all together in Langfuse Academy.

Easy to get started and integrate

You pick the models and frameworks. We'll fit in.

• Our free Cloud plan includes 50,000 usage units a month, no credit card required.
• Our Cloud pricing is public. Paid plans include unlimited users; Pro and Enterprise include three years of history.
• Start with our Python and JavaScript/TypeScript SDKs or 100+ integrations across frameworks, model providers, gateways, and tools.
• Native OpenTelemetry support lets you reuse your tracing instrumentation.
• Prefer to self-host? Start locally with Docker Compose, or deploy on your infrastructure with Helm and Terraform templates.

At Hugging Face, Niels Rogge was up and running in minutes. Khan Academy used our API to build a client for its Go services.

Developer-first, down to the details

We care about making Langfuse enjoyable to use every day.

• Review answers, test prompts, and run evals together in the UI.
• Access your data and build custom workflows with our public APIs and SDKs.
• Your coding agent is welcome, too: use our Agent Skill, CLI, or MCP server.
• Ask the Assistant to investigate traces or create datasets and dashboards. It asks before making changes (Cloud; self-hosted public beta).

Ravenna chose Langfuse for its clean, well-built product. At Canva, a marketing manager built a tone-of-voice evaluator without engineering help.

Security and compliance

• We take security and compliance seriously.
• Certifications: Langfuse Cloud has completed a SOC 2 Type II audit and is ISO 27001 certified. We also conduct annual third-party penetration tests.
• Privacy: GDPR compliant with a DPA available. For HIPAA workloads, eligible Pro and Enterprise customers can use a dedicated cloud region with a signed Business Associate Agreement (BAA) in place before processing PHI.
• Data regions: Choose between US, EU, Japan, or HIPAA-ready Cloud regions, or self-host in your own infrastructure.
• Hardening for Government (in development): FIPS-compatible deployment support, hardened images, and guidance for operating the full stack in air-gapped government environments are under development for Self-Hosted Enterprise.
• Data control: Data masking, data retention, and data deletion, plus Cloud encryption at rest and in transit. Configurable retention depends on plan and deployment.
• Enterprise access: SSO, fine-grained roles, SCIM provisioning, and audit logs, with availability depending on plan and deployment.
• More details in our Security & Privacy Center.

A partner you can build on

• We're part of ClickHouse, working directly with its engineers on performance, reliability, and security.
• We follow semantic versioning and run SDK integration and end-to-end tests in CI.
• We publish compatibility policies and migration paths. Self-hosters choose when to upgrade.
• Enterprise plans include uptime and support SLAs.

Thoughtworks' Technology Radar has featured Langfuse three times, most recently in April 2026. Our adoption and community metrics are public, too.

We build with our users

You can talk directly to the team building Langfuse about your setup, ideas, or where you're stuck.

The Canva team said: "Other vendors came back with Figma prototypes. Meanwhile, Langfuse shipped two features. That's when we knew."

Get started

Comparison to other platforms

Still weighing your options? Our comparisons cover LangSmith, Braintrust, Arize / Phoenix, and Datadog.

Public metrics

---

---