- CollectionObservability & Tracing
- Classificationagent-native
- LicenseOpen-source (MIT / enterprise)
- MCPAvailable
- Agent SkillAvailable
- InterfacesAgent Skill · MCP · SDK · API
What it does
Langfuse provides structured tracing for multi-step agent workflows. It captures the full execution trajectory of an agent — LLM calls, tool invocations, retriever steps, guardrail checks — organized in a typed, hierarchical trace structure. Production traces can be converted to evaluation datasets with one click, and OpenTelemetry integration allows existing monitoring infrastructure to receive agent telemetry.
Agent-facing interfaces
- Catalog protocols: Agent Skill, MCP, SDK, API
- MCP: ✅ Available
- Agent Skill: ✅ Official skill published at langfuse/skills — listed on skills.sh/langfuse/skills
- Onboarding type:
agent-skill
Protocol surface
| Interface | Detail |
|---|---|
| Python SDK | langfuse.trace(), @observe() decorator, context manager |
| TypeScript SDK | Equivalent TypeScript API |
| OpenTelemetry | OTEL-compatible; works with any OTEL-instrumented agent framework |
| OpenAI Agents SDK | Native auto-instrumentation |
| Claude SDK | Native auto-instrumentation |
| LangChain | Native callback integration |
| LangGraph | Native integration |