Langfuse
by Langfuse (acquired by ClickHouse, Inc. in January 2026)
Open source LLM engineering platform
Langfuse is an open-source LLM engineering and observability platform for tracing, evaluating, debugging, and improving AI applications, with both managed cloud and self-hosted deployment options.
At a Glance
- Category
- Developer Tools
- Pricing
- freemium, subscription, usage-based, open-source
- Target Market
- Startups, Enterprises, Individual developers, AI engineering teams
- Founded
- 2022
- Headquarters
- Berlin, Germany
- Customers
- Trusted by 19 of the Fortune 50 and 63 of the Fortune 500 (per ClickHouse acquisition announcement)
Key Features
- ✓LLM Observability & Tracing
Hierarchical traces capture every LLM call, tool invocation, retrieval, and embedding step with inputs, outputs, timing, cost, latency, and metadata, filterable by user, session, cost, latency, or custom metadata.
- ✓Prompt Management
Version, manage, and deploy prompts separately from code with one-click deployments and rollbacks and strong server- and client-side caching to minimize latency.
- ✓Evaluations
Supports LLM-as-a-judge, code-based evaluators, user feedback collection, manual labeling, and custom evaluation pipelines via APIs and SDKs.
- ✓Datasets & Experiments
Build test sets and benchmarks to run structured experiments, enable pre-deployment testing, and continuously improve LLM applications.
- ✓Playground
Interactive interface for testing and iterating on prompts and model configurations, with the ability to jump from a bad trace directly into the playground.
Capabilities
Use Cases
- •Debugging multi-step LLM and agent pipelines
Use hierarchical trace visualization to inspect retrieval, reranking, tool calls, and generation steps separately, reducing debug time in complex chains and agents.
- •Monitoring production LLM cost, latency, and quality
Track token usage, cost, latency, and quality across LLM applications in production, watch for issues like hallucinations, and analyze request history when things go wrong.
- •Evaluating and iterating on prompts before deployment
Manage prompt versions, run experiments against datasets, and apply LLM-as-a-judge and code evaluators to test changes before shipping to production.
Ideal For
Best For
- ✓AI/ML engineering teams building LLM applications
- ✓Developers debugging multi-step LLM and agent pipelines
- ✓Teams needing prompt management and evaluation workflows
- ✓Organizations requiring self-hosted LLM observability
Integrations
Deployment
Market & Ratings
Trusted by 19 of the Fortune 50 and 63 of the Fortune 500 (per ClickHouse acquisition announcement)
Market Analysis
Pros
- ✓Open-source with free, uncapped self-hosting
- ✓Detailed hierarchical tracing for debugging complex/multi-step LLM and agent pipelines
- ✓Clean prompt versioning and experiment workflows
- ✓Straightforward SDK setup and broad framework integrations
Cons
- ✗Bulk trace querying can become slow at high traffic volumes
- ✗UI can get crowded on complex multi-step traces and agents
- ✗Self-hosting total cost of ownership can be high for teams not already running ClickHouse/Postgres at scale
Pricing
Hobby
$0/month
- ✓50k units/month included
- ✓30 days data access
- ✓2 users
- ✓Community support
- ✓No credit card required
- ✓All core features with limits
Core
$29/month
- ✓100k units/month included, additional $8/100k units
- ✓90 days data access
- ✓Unlimited users
- ✓In-app support
Pro
$199/month
- ✓100k units/month included, additional $8/100k units
- ✓3 years data access
- ✓Unlimited annotation queues
- ✓High rate limits
- ✓SOC2 & ISO27001 reports, HIPAA-compliant BAA available
- ✓Optional Teams add-on $300/month (SSO, RBAC, dedicated Slack)
Enterprise
$2,499/month
- ✓Everything in Pro + Teams
- ✓Custom rate limits
- ✓Uptime and support SLA
- ✓Dedicated support engineer
- ✓Optional yearly commitment with custom volume pricing
Self-Hosted (Open Source)
$0
- ✓Self-host all core features for free
- ✓Open source, no hard cap on events or traces
- ✓Deploy via Docker, Kubernetes, or cloud templates
Cloud usage is metered in 'units' with graduated overage rates decreasing from $8/100k units (100k-1M) to $6/100k units (50M+). Paid cloud tiers include unlimited users (no per-seat charge). Self-hosting the open-source version is free of license cost, though infrastructure (ClickHouse, Postgres) must be operated by the user.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
Subscribe