An open-source (MIT) LLM engineering platform for call tracing, cost, prompt management and evaluation of LLM apps in one place
Information updated 2026-10-20, source: official site and public reports. Prices and benefits may change. Check the official site before you pay.
Pros
- MIT open source/free self-hosting (all core features)
- Tracing/token/cost aggregation
- Prompt version management/playground
- LLM-as-judge/dataset evaluation
- Free cloud plan
Cons
- A developer tool, so hard for non-developers
- The cloud charges extra when units are exceeded
- No Korean interface
These are opinions compiled by Pickx from public sources.
Cautions
The usage unit is 'units' (traces, observations and so on), with overage charged at $8 per 100k. SSO/RBAC need the Teams add-on ($300 per month) or Enterprise. ClickHouse acquired it, and it was officially announced that it will remain open source with no plans to change the license
At a glance
Langfuse is an open-source platform that tracks which prompts an LLM app used to call which models and how much cost/latency there was, and also handles prompt management and quality evaluation. It suits development teams running AI features in real services. You can use it in the cloud (a free plan exists) or self-hosted (MIT, free).
Main features/supported AI
| Feature/model | Description | Notes |
|---|---|---|
| Tracing | Agent traces/graphs, per session/user, token/cost tracking, multimodal | |
| Prompt Management | Prompt versioning, deployment, caching, playground | |
| Evaluation | Online/offline evaluation, datasets/experiments, LLM-as-judge, human annotation queues | |
| Metrics | Custom dashboards/alerts | Langfuse v4+ |
| Integrations | Python/JS SDK, OpenTelemetry, LiteLLM proxy logging, public API | Works with any model provider |
Plans
| Plan | Price (monthly billing / yearly billing) | What you can do/limits | Features |
|---|---|---|---|
| Hobby (cloud) | Free | 50,000 units per month, 30 days of data, 2 users | For personal projects |
| Core |
| 100,000 units per month included ($8 per extra 100,000), 90 days of data, unlimited users | |
| Pro |
| 100,000 units per month included, 3 years of data/retention policy management | Teams add-on $300/mo (SSO/RBAC) |
| Enterprise (cloud) |
| 100,000 units per month included, 3 years of data | Custom volume pricing |
| Self-hosted Open Source | Free | All core features, unlimited projects/users | MIT license |
| Self-hosted Enterprise | Quote | Admin API, project RBAC, retention policies, audit logs, SCIM, dedicated support | SOC 2/ISO 27001 reports |
Who it suits
- Hobby: Developers who want to record the calls/costs of a personal AI project for free
- Core: When a small team wants to manage the cost/quality of AI features in operation
- Self-hosted Open Source: Companies that cannot send log data outside
Compared with similar services
- Langfuse vs LangSmith: LangSmith is a commercial service centered on the LangChain ecosystem, while Langfuse is MIT open source and free to self-host
- Langfuse vs LiteLLM: LiteLLM is a gateway that relays/controls model calls, while Langfuse is an observability tool that records/analyzes/evaluates calls (they can be used together)
- Langfuse vs Helicone: Both are LLM observability tools, and Langfuse has a wider range of prompt management/evaluation features
Notes
- Information checked: 2026-10-20
- Langfuse official pricing (cloud)
- Langfuse self-hosting pricing
- Langfuse joins ClickHouse (official blog)
Good for
0 views
Prices converted at 1 USD = ₩1,329.49 (2026-10-19). Original price shown in brackets. Taxes as listed by each provider.
Comments
Sign in to leave a comment.
No comments yet. Be the first.
Your comments and nickname are public. If you see abuse or advertising, report it here and we will review and remove it.