LLM Monitoring for Every Framework You Use
Fluiq auto-instruments every major LLM provider, agent framework, and vector database with two lines of Python, zero decorators, zero wrappers, zero code changes to your existing code.
OpenAI, Anthropic, Gemini, Vertex AI, every provider call traced with token counts and USD cost.
OpenAI
Two lines of Python give every OpenAI call, chat completions, streaming, embeddings, images, audio, a full trace with token counts, USD cost at OpenAI rates, and security scanning.
Anthropic
Auto-instrument every Claude API call, Messages, streaming, and the Beta client, with zero code changes.
Google Gemini
Auto-instrument every Gemini generate_content call, streaming and non-streaming, with zero code changes.
Google Vertex AI
Auto-instrument enterprise Vertex AI model calls with zero code changes.
LangChain, LangGraph, CrewAI, Google ADK, MCP, full agent span trees.
LangChain
Auto-instrument every LangChain chain step, LLM calls, tool use, retriever queries, with zero code changes.
LangGraph
Auto-instrument every LangGraph graph execution, node by node, edge by edge.
CrewAI
Auto-instrument every CrewAI crew execution, agent by agent, task by task.
Google ADK
Auto-instrument every Google Agent Development Kit run with zero code changes.
MCP
Auto-instrument every Model Context Protocol tool call, list_tools, call_tool, and read_resource, with zero code changes.
Pinecone, Chroma, Weaviate, FAISS, Qdrant, query latency and RAG pipeline child spans.
Pinecone
Auto-instrument every Pinecone index operation, query, upsert, fetch, delete, with zero code changes.
Chroma
Auto-instrument every ChromaDB collection operation, query, add, update, get, with zero code changes.
Weaviate
Auto-instrument every Weaviate query and insert, near-text, near-vector, batch operations, with zero code changes.
FAISS
Auto-instrument every FAISS index operation, search, add, and batch queries, with zero code changes.
Qdrant
Auto-instrument every Qdrant search, upsert, and scroll call, with zero code changes.
Don't see your framework?
Use the @trace decorator to manually instrument any Python function. Or reach out, new integrations ship regularly.