Langfuse vs LangSmith
Side-by-side comparison from the Agent Observability Index: licensing, self-hosting, pricing model and integrations — no vendor copy, primary sources linked.
| Langfuse | LangSmith | |
|---|---|---|
| One-liner | Open-source (MIT) LLM engineering platform combining tracing, prompt management, evals and datasets, widely used as the default self-hosted observability stack. | Closed-source tracing, evals and monitoring platform from the LangChain team, deepest integration with LangChain/LangGraph but usable via OTel from any stack. |
| Category | Observability & Tracing | Observability & Tracing |
| Open source | Yes | No |
| Self-hostable | Yes | Yes |
| Pricing model | freemium | freemium |
| Pricing notes | Free Hobby tier (50k units/mo); Core $29/mo, Pro $199/mo, unlimited users on paid tiers; MIT core fully self-hostable | Free developer seat; Plus ~$39/seat/mo plus usage-based trace billing; self-hosting is enterprise-only |
| Frameworks | langchain, openai-sdk, llamaindex, litellm, opentelemetry, vercel-ai, haystack | langchain, langgraph, openai-sdk, opentelemetry |
| Funding / ownership | YC W23; $4M seed (Lightspeed, La Famiglia) | LangChain raised $125M Series B (2025) at ~$1.25B valuation |
How to choose
- Choose Langfuse if open-source licensing or full data control is a hard requirement.
- Both link to primary pricing sources below — verify current tiers before committing; this market shifts monthly.