What is
Langfuse
?
Most platform initiatives fail the same way: they start too big, get funded like projects instead of products, build for too long without testing with users, and confuse renaming an infrastructure team with building a platform team. Instead, start small, learn fast and scale deliberately.
At a glance
Attribute | Value |
|---|---|
Primary use case | Trace, evaluate and improve AI agents and LLM applications |
Deployment type | Hosted cloud platform or self-hosted via Docker Compose, Kubernetes Helm, or Terraform on AWS, GCP and Azure |
Pricing model | Free/open-source self-hosting, free cloud entry option, paid cloud plans with usage-based pricing and enterprise plans |
API compatibility | OpenTelemetry (OTel), REST APIs, MCP, native Python and TypeScript SDKs |
Supported integrations | LangChain, Vercel AI SDK, LiteLLM, Pydantic AI, Google ADK, CrewAI, LiveKit, OpenAI, Anthropic, Amazon Bedrock, Azure OpenAI, Google Gemini, LlamaIndex, OpenAI Agents SDK, Dify, Langflow, OpenRouter, n8n, Spring AI, Promptfoo, Ragas, AutoGen |
Source type | Open |
Cloud provider | Multi-Cloud |
Ideal team size | Small teams to enterprise |
Required skills | SDK/API integration, OpenTelemetry instrumentation, familiarity with LLM applications and execution traces; Docker, Kubernetes or Terraform for self-hosting |
AEP branch | Agent Infrastructure |
AEP category | Governance |
AEP subcategory | Agent Observability |
AEP detail | Traces & Telemetry |
Profile
Traces & Telemetry is Langfuse’s most specific AEP level. Langfuse captures hierarchical traces of LLM calls, tool invocations, and retrieval steps, and connects trace data with monitoring, evaluation, prompt management, datasets, and experiments.
Teams can filter traces by user, session, cost, latency, or custom metadata. The platform also supports evaluating production data, testing prompts against production inputs, comparing models, and collecting human feedback.
Focus
Langfuse is intended for teams building and operating AI agents and LLM applications that need to inspect execution behavior and improve quality, cost, and latency. It addresses the need to connect production traces with evaluation, experiments, prompt changes, and human review.
Background
The supplied company-background research identifies Langfuse as a YC W23 company and reports that it is part of ClickHouse, Inc. as of 2026. The research does not provide a founding date or funding details.
Langfuse’s site describes the platform as open source under the MIT license and reports more than 22,000 GitHub stars, an active open-source community, and weekly releases. It also highlights APIs and exports, including REST APIs, a query SDK, and S3 blob storage export.
Features
Observability and tracing
Inspect the execution of LLM applications and agents, including cost and latency for individual calls. Langfuse uses hierarchical traces to capture LLM calls, tool invocations, and retrieval steps, with filters for user, session, cost, latency, and custom metadata.
Evaluation
Evaluate production behavior or experiment results to assess application quality. Langfuse supports LLM-as-a-judge, heuristic functions, and human review, and can attach evaluations to production data or experiments.
Prompt management
Manage prompts separately from application code and deploy or roll back prompt changes. Langfuse provides prompt management with one-click deployments and rollbacks, and can link prompts to tracing.
Playground and experiments
Test prompt and model changes against real production inputs and compare outcomes. The Playground supports side-by-side model comparisons, while experiments use defined test cases to compare results.
Human annotation and datasets
Enable collaborative review of traces and build golden datasets from human feedback. Langfuse provides human-in-the-loop workflows for trace review and dataset creation.
Cost and latency monitoring
Track application costs, latency, and quality to help teams identify areas to investigate. Dashboards display these measures, and automated alerts can report on them.
Integrations and instrumentation
Connect Langfuse with agent frameworks, model providers, and other tools without framework lock-in. The platform supports 100+ integrations, works with languages that support OpenTelemetry instrumentation, and offers native Python and TypeScript SDKs.
Pros and cons
Pros
Combines tracing, prompt management, evaluation, datasets, experiments, and human feedback in one platform.
Supports OpenTelemetry, REST APIs, MCP, and native Python and TypeScript SDKs; the site lists more than 100 integrations.
Offers hosted cloud and self-hosting options, with Docker Compose, Kubernetes Helm, and Terraform deployment paths for AWS, GCP, and Azure.
The site states that all product features are MIT licensed and that Langfuse scales to billions of monthly events.
Cons
Native SDKs are listed for Python and TypeScript; instrumentation for other languages depends on OpenTelemetry support.
Self-hosting requires deploying Langfuse with Docker Compose, Kubernetes Helm, or Terraform and operating the selected environment, unlike using the hosted cloud platform.
FAQs
Frequently asked questions about Langfuse
How does Langfuse trace AI applications?
Langfuse captures hierarchical traces of LLM calls, tool invocations, and retrieval steps, and connects trace data with monitoring, evaluation, prompt management, datasets, and experiments.
What does Langfuse integrate with?
Langfuse supports OpenTelemetry (OTel), REST APIs, MCP, native Python and TypeScript SDKs, and integrations including LangChain, Vercel AI SDK, LiteLLM, Pydantic AI, Google ADK, CrewAI, LiveKit, OpenAI, Anthropic, Amazon Bedrock, Azure OpenAI, Google Gemini, and LlamaIndex.
What does Langfuse cost?
Langfuse pricing is Free/open-source self-hosting, free cloud entry option, paid cloud plans with usage-based pricing and enterprise plans.
What skills are required to instrument Langfuse?
Langfuse requires SDK/API integration, OpenTelemetry instrumentation, familiarity with LLM applications and execution traces; Docker, Kubernetes or Terraform for self-hosting.
What are Langfuse's limitations?
Langfuse lists native SDKs for Python and TypeScript; instrumentation for other languages depends on OpenTelemetry support. Self-hosting requires deploying Langfuse with Docker Compose, Kubernetes Helm, or Terraform and operating the selected environment.
