langflow-ai/langflow

Langflow is a visual drag-and-drop platform for building AI agents and RAG pipelines. You connect LLMs, vector databases, tools, memory modules, and data sources as nodes on a canvas — then deploy the entire workflow as an API endpoint with one click. The 147K GitHub stars tell part of the story. The real draw is that Langflow collapses a multi-day coding task into a few hours of visual authoring. You build a RAG pipeline by dragging a document loader, an embedder, a vector store, and a retriever into a flow. Connect them. Hit deploy. Your pipeline is live as a REST API. Under the hood, everything runs on LangChain. Langflow is essentially a visual IDE for LangChain's component library — every chain, agent, tool, and memory type LangChain supports is available as a draggable node. Python developers can extend it with custom components. TypeScript is used for the frontend (25.9% of the codebase). Key capabilities that separate Langflow from alternatives like Dify and n8n: Built-in MCP server: Every flow automatically becomes available as an MCP tool. Your Claude or Cursor agent can call any Langflow workflow directly. Model-agnostic: OpenAI, Anthropic, Google, Ollama, Hugging Face — swap models without changing your flow logic. Production API: Not just a prototyping tool. Flows deploy as production-ready APIs with authentication, rate limiting, and monitoring. Version 1.8.3 (March 26, 2026) is the latest release. Langflow Cloud starts at $25/month. Self-hosted is free and open-source under the MIT license. If you are building AI agent pipelines visually, also check out OpenBox AI for runtime governance of your deployed agents, and MetaMCP for managing MCP server infrastructure.