Langfuse · LLM observability platform

Langfuse

Open-source tracing, prompt management and evaluation for LLM applications, self-hosted or in the cloud.

Provider
Langfuse
Latest confirmed model
Not applicable / not verified
Pricing
Open source
Free plan
Yes
API
Yes · free quota
Status
Generally available
Last verified
Oct 19, 2026

What is Langfuse?

Langfuse records what your LLM app does, including model calls, retrieval steps and agent actions, so you can debug, measure cost and quality, and improve prompts. It also offers prompt versioning, datasets, evaluations and a playground. The core is MIT-licensed and free to self-host; Langfuse Cloud is a hosted version with usage-based plans.

Understanding and improving LLM app behaviour, cost and quality in production.

Why use Langfuse?

Practical benefits

  • MIT-licensed core, free self-hosting
  • Free cloud Hobby plan without a credit card
  • Integrations with LangChain, LlamaIndex, OpenAI, CrewAI and more
  • Typed Python and JS/TS SDKs

Who it suits

  • Teams running LLM apps in production
  • Developers evaluating prompts
  • Companies needing self-hosted observability

When another tool may fit better

  • Single-user experiments with no need for tracing
  • Teams wanting all enterprise features free when self-hosting (ee folders are excluded from MIT)
Learning curve
Some experience helps
Skills needed
Programming; DevOps for self-hosting.

Example workflows

  1. Add the Langfuse SDK, trace each request, review traces and scores in the dashboard

What can it do?

  • Tracing and observabilityTraces LLM calls, retrieval and agent actions.
  • Prompt managementVersioned prompts.
  • EvaluationsLLM-as-a-judge, code evaluators, user feedback and manual labelling.
  • Datasets and playgroundTest sets and prompt iteration.
  • Self-hostingDocker Compose, Kubernetes (Helm) and Terraform templates for AWS, Azure, GCP.
  • Data retention management Depends on planAvailable on Pro and Enterprise cloud plans.

Good for

  • Debugging agents
  • Tracking LLM cost and latency
  • Prompt A/B testing
  • Quality evaluation

Where it runs

Pricing and plans

Open source Self-hosting the open-source core is free. Langfuse Cloud: Hobby free (50k units/month), Core $29/month, Pro $199/month and Enterprise $2,499/month, each with 100k units included and graduated overage from $8 per 100k units.

PlanPriceBillingFree trial / tierUsage limitsMain inclusionsModel accessAPI accessRestrictions
Self-hosted (open source)Freeper month Free tierYour infrastructure
  • Core platform
—YesEnterprise features need a licence
HobbyFreeper month Free tier50k units/month; 30-day data access; 2 users
  • Tracing
  • Prompts
  • Evals
—YesNo overage
Core$29/moper month —100k units included; $8 per 100k overage (lower at volume); 90-day data access
  • Unlimited users
—Yes—
Pro$199/mo
Teams add-on $300/month
per month —100k units included; 3-year data access
  • Data retention management
  • Unlimited users
—Yes—
Enterprise$2,499/mo
Custom yearly pricing also available
per month —100k units included; 3-year data access
  • Data retention management
  • Unlimited users
—Yes—

Prices in USD · Global list price · Taxes may apply · Verified Oct 19, 2026

Official pricing page (opens in a new tab)

API and developer access

Official API
Yes
Authentication
API keys (public/secret)
Pricing model
per 100k units
Free API quota
Hobby: 50k units per month
Rate limits
Not documented
Context / input limits
Not documented
SDKs
Python, JavaScript/TypeScript
Streaming
Not documented
Batch processing
Not documented
Structured output
Not documented
Tool / function calling
Not documented
Commercial use
MIT except ee folders
Data retention & privacy
Data access 30 days (Hobby), 90 days (Core), 3 years (Pro/Enterprise)

Published API rates

ItemPriceUnit
Overage 100k-1M units$8per 100k units
Overage 1M-10M units$7per 100k units
Overage above 50M units$6per 100k units

A free Langfuse account is not the same thing as free API access. The API allowance above is the provider's documented developer quota.

Verified Oct 19, 2026

Advantages and disadvantages

Advantages

  • Open source with full self-hosting
  • Broad integrations
  • Transparent usage pricing

Disadvantages

  • Enterprise (ee) features are outside the MIT licence
  • Hobby plan limited to 2 users and 30-day data access
  • Self-hosting requires infrastructure work

Privacy: Self-hosting keeps all trace data on your infrastructure.

Real-world examples

A solo developer tracing a chatbot

  1. Sign up for the free Hobby plan (no card)
  2. Add the SDK to the app
  3. Inspect up to 50k units per month

free plan works

A company self-hosting for privacy

  1. Deploy with the Helm chart on Kubernetes
  2. Point the SDK at your instance
  3. Keep all traces in-house

free plan works

A team evaluating prompt changes

  1. Create a dataset of test inputs
  2. Run two prompt versions
  3. Score with LLM-as-a-judge and compare

free plan works

Langfuse alternatives

LangChain

Open-source framework for building LLM apps and agents that work with many model providers.

has a free plan · offers an API · open source / open-weight

Langfuse vs LangChain

LlamaIndex

Open-source framework for connecting LLMs to your data, plus the commercial LlamaParse document platform.

has a free plan · offers an API · open source / open-weight

Langfuse vs LlamaIndex

CrewAI

Open-source Python framework for teams of role-based AI agents, with a hosted platform for building and running them.

has a free plan · offers an API · open source / open-weight

Langfuse vs CrewAI

Compare Langfuse with these side by side

Frequently asked questions

Is Langfuse free?

Yes. The open-source core can be self-hosted for free, and Langfuse Cloud has a free Hobby plan with 50k units per month and no credit card.

What licence does Langfuse use?

MIT, except for code in the ee folders, which is excluded from the MIT licence.

How much does Langfuse Cloud cost?

Core is $29/month, Pro $199/month and Enterprise $2,499/month, each including 100k units, with overage starting at $8 per 100k units.

How do I self-host Langfuse?

Use Docker Compose locally or on a VM, Helm on Kubernetes for production, or Terraform templates for AWS, Azure and GCP.

Does Langfuse work with LangChain and LlamaIndex?

Yes, both are listed among its core integrations, along with OpenAI, LiteLLM, Vercel AI SDK and others.

How long is data kept?

Cloud data access is 30 days on Hobby, 90 days on Core and 3 years on Pro and Enterprise.