Integrations
Connect Muster to the LLM frameworks, model providers, gateways, and tools you already use — 100+ integrations covered.
Muster is built on top of the open-source Langfuse core, which means it speaks the same OpenTelemetry-based ingestion API and works with the same SDKs and integrations the wider Langfuse ecosystem supports. If your stack already emits traces to Langfuse, pointing them at Muster takes a host swap and an API key — nothing else changes.
You can instrument your application in three ways:
- Native SDKs — install the
langfusePython or JS/TS package and either wrap your LLM calls with the provided decorators / clients, or push spans manually. This is the path with the richest tooling (prompt management, evaluations, datasets) and works against any model provider. - OpenTelemetry endpoint — point any OTel-compatible tracer (Java, Go, Rust, .NET, etc.) at the Muster OTLP endpoint and your spans show up as traces. Use this when there is no first-party SDK for your runtime.
- Drop-in framework integrations — frameworks like LangChain, LlamaIndex, the Vercel AI SDK, or LiteLLM ship Muster/Langfuse callbacks out of the box. Wire the callback once and every call routed through the framework is captured automatically.
The pages below cover every integration grouped by category. Pick the one that matches what you're already using; if you don't see a fit, the OpenTelemetry path covers everything else.
Native
- OpenTelemetry — language-agnostic OTLP ingestion. The lowest common denominator, works from anywhere.
Frameworks
- LangChain ecosystem — LangChain · DeepAgents · LangServe
- LlamaIndex — LlamaIndex · Workflows
- Agent SDKs — OpenAI Agents SDK · Claude Agent SDK (Python) · Claude Agent SDK (JS/TS) · Google ADK · CrewAI · AutoGen · Pydantic AI · Semantic Kernel · Microsoft Agent Framework
- JS/TS frameworks — Vercel AI SDK · Mastra · VoltAgent
- Voice & multimodal — Pipecat · LiveKit Agents
- Specialised — Haystack · DSPy · Instructor · Mirascope · Ragas · smolagents · Agno · BeeAI · Strands Agents · Amazon AgentCore · watsonx Orchestrate
- JVM — Spring AI · Quarkus + LangChain4j · Embabel · Koog (Kotlin)
- Other languages — Swiftide (Rust)
- Durable execution — Temporal · Restate
- SDK-style gateways — LiteLLM SDK
Model Providers
- OpenAI / Azure — OpenAI (Python) · OpenAI (JS/TS) · Assistants API
- Anthropic — Python · JS/TS
- AWS Bedrock — Bedrock · Bedrock Agents
- Google — Gemini · Vertex AI
- Frontier providers — Cohere · Mistral SDK · Hugging Face · Groq · DeepSeek · Together.ai · xAI Grok
- Hosted inference — Databricks · Fireworks AI · Cerebras · Baseten · Novita AI · BytePlus · CometAPI · Qwen
- Self-hosted — vLLM · Ollama
- Evaluation — Cleanlab TLM
Gateways
- LiteLLM Proxy · OpenRouter · Portkey · Helicone · Vercel AI Gateway · TrueFoundry · Kong AI Plugin · Anannas
No-Code
Analytics
- PostHog · Mixpanel · Trubrics · Coval · Agentic Data Stack
Other
- AI coding tools — Claude Code · Cursor · Kiro
- Chat UIs — LibreChat · Gradio
- Tools & utilities — Firecrawl · Exa · Milvus · Promptfoo · mcp-use · Cognee · Parallel AI · Inferable · Testable Minds · Weco · Zapier · OpenClaw
Don't see your tool? The OpenTelemetry path covers everything else — point any OTLP-compatible tracer at Muster and you're done.