// the find
FlowiseAI/Flowise
Build AI Agents, Visually
Flowise is a drag-and-drop UI for building LLM workflows and agents, primarily on top of LangChain. It targets non-developers and prototypers who want to wire together RAG pipelines, chatbots, and multi-agent systems without writing code. The visual canvas outputs a JSON flow definition that the Node.js backend executes.
The component library is genuinely large — 200+ integrations covering most LLM providers, vector stores, and tools, so you're unlikely to hit a missing connector for common use cases. The monorepo structure is clean: agentflow (the canvas React component) is published as a standalone npm package, which means you can embed just the visual editor in your own app. Active development cadence with daily commits and a proper CI pipeline including load tests. Docker Compose setup is straightforward and production-usable without modification.
LangChain dependency is a liability — LangChain's frequent breaking changes have caused real pain for Flowise users, and the abstraction leaks constantly when debugging why a chain behaves unexpectedly. Visual flows do not version-control well; the exported JSON is large, diff-unfriendly, and merging two people's changes to the same flow is a manual nightmare. The execution model is opaque: when a flow fails in production, tracing which node caused the issue requires digging through logs rather than anything the UI exposes. No first-class support for testing a flow against a fixture dataset — you can only run it live.