Flux0 – LLM-framework agnostic infra for LangChain agents with streaming, sessions, and multi-agent support

[asaf]

Hey guys,

We built Flux0, an open framework that lets you build LangChain (or LangGraph) agents with real-time streaming (JSONPatch over SSE), full session context, multi-agent support, and event routing — all without locking you into a specific agent framework.

It’s designed to be the glue around your agent logic:

🧠 Full session and agent modeling

📡 Real-time UI updates (JSONPatch over SSE)

🔁 Multi-agent orchestration and streaming

🧩 Pluggable LLM execution (LangChain, LangGraph, or your own async Python code)

You write the agent logic, and Flux0 handles the surrounding infrastructure: context management, background tasks, streaming output, and persistent sessions.

Think of it as your backend infrastructure for LLM agents — modular, framework-agnostic, and ready to deploy.

→ GitHub: https://github.com/flux0-ai/flux0

Would love feedback from anyone building with LangChain, LangGraph, or exploring multi-agent setups!