Langfuse

Langfuse is an open source AI engineering platform for tracing, evaluating, monitoring, and improving LLM applications and AI agents.

About Langfuse

Langfuse is an open source AI engineering platform for building, evaluating, and improving LLM applications and AI agents. It combines observability, evaluation, prompt management, datasets, experiments, monitoring, and analytics in one platform.

Langfuse captures traces from LLM calls, tool invocations, retrieval steps, agent graphs, and related application workflows. Teams can inspect production behavior, monitor cost and latency, compare experiments, evaluate outputs, manage prompts, and investigate application issues.

The platform supports automated evaluations using LLM based judges and code evaluators, as well as human review through annotation queues. Developers can use Langfuse Cloud or self host the platform, connect through SDKs and APIs, and integrate it with frameworks, model providers, gateways, developer tools, and OpenTelemetry based systems.

Key Features

Agent and LLM Observability: Capture hierarchical traces for LLM calls, tool invocations, retrieval steps, sessions, users, and agent graphs.
AI Evaluation: Evaluate outputs using LLM based judges, heuristic functions, code evaluators, experiments, datasets, and human review.
Prompt Management: Create, version, test, deploy, and roll back prompts separately from application code.
Production Monitoring: Track model usage, token consumption, cost, latency, quality signals, and custom metadata.
Experiments and Datasets: Define test cases, run experiments, and compare model, prompt, or workflow changes.
Developer and Platform Access: Use Python and JavaScript SDKs, OpenTelemetry, REST APIs, a CLI, and an MCP server.

Pricing

Hobby: Free. Includes 50,000 units per month, 30 days of data access, two users, all platform features with limits, and community support through GitHub.
Core: $29/month. Includes 100,000 units per month, 90 days of data access, unlimited users, and in-app support. Additional usage is billed at $8 per 100,000 units at the listed tier.
Pro: $199/month. Includes 100,000 units per month, three years of data access, data retention management, unlimited annotation queues, high rate limits, prioritized in-app support, and access to SOC 2 Type II and ISO 27001 reports. Additional usage is billed at $8 per 100,000 units at the listed tier.
Teams Add-on: $300/month. Available as an optional add-on for the Pro plan and includes enterprise SSO, SSO enforcement, fine grained RBAC, and support through a dedicated Slack channel.
Enterprise: $2,499/month. Includes 100,000 units per month, audit logs, SCIM API, custom rate limits, uptime SLA, support SLA, and a named lead support engineer. Additional usage is billed at $8 per 100,000 units at the listed tier.
Self-Hosted: Core Langfuse features can be self hosted without charge. Some add-on features require a license key.

Pricing last updated: September 29, 2026 at 12:00 AM

Use Cases

Monitoring and debugging production AI agents
Evaluating LLM applications, retrieval systems, and agent workflows
Managing and testing prompts across development and production
Tracking model usage, token consumption, cost, and latency
Building datasets and running experiments to compare AI changes
Reviewing AI outputs through human annotation workflows

Pros & Cons

Pros:

Open source platform with MIT licensed product features
Supports tracing, evaluations, prompt management, experiments, and monitoring in one platform
Works with more than 100 integrations and supports OpenTelemetry
Provides cloud and self hosted deployment options
Offers SDK, REST API, CLI, and MCP server access

Cons:

Cloud plans include usage, data retention, rate, user, or support limits depending on the plan
The Hobby plan includes 50,000 units per month, 30 days of data access, and two users
Some self hosted add-on features require a license key

Integrations

OpenTelemetry, LangChain, Vercel AI SDK, LiteLLM, Pydantic AI, Google ADK, CrewAI, LiveKit, OpenAI, Anthropic, Amazon Bedrock, Azure OpenAI, Mistral AI, Google Gemini, xAI, vLLM, Groq, Claude Code, Cursor, Codex, OpenWebUI, Dify, n8n, PostHog, Python, JavaScript, TypeScript, Go, Java, .NET, Ruby, PHP, Swift

FAQ

What is Langfuse?

Langfuse is an open source AI engineering platform for observing, evaluating, and improving LLM applications and AI agents.

What can Langfuse observe?

Langfuse can capture traces for LLM calls, tool invocations, retrieval steps, sessions, users, agent graphs, prompts, outputs, cost, latency, and related application metadata.

Does Langfuse support AI evaluations?

Yes. Langfuse supports LLM based judges, heuristic functions, code evaluators, datasets, experiments, and human review through annotation queues.

Can Langfuse manage prompts?

Yes. Langfuse provides prompt management features for creating, versioning, testing, deploying, and rolling back prompts separately from application code.

Does Langfuse support self hosting?

Yes. Langfuse can be self hosted using its open source deployment options. Some add-on features require a license key.

Does Langfuse have a free plan?

Yes. The Hobby cloud plan is free and includes 50,000 units per month, 30 days of data access, two users, and community support through GitHub.

Which integrations does Langfuse support?

Langfuse supports more than 100 integrations across frameworks, model providers, gateways, no code tools, analytics platforms, developer tools, and OpenTelemetry based systems.

Categories:

AI Observability Platforms Dev Tools