Langflow

Langflow is a low-code, node-based canvas for assembling LLM agents and retrieval-augmented generation pipelines. You drag components onto a graph, wire prompts to models to tools to vector stores, and iterate on the flow live before exporting it as an API endpoint or a Python module. It supports the usual LLM providers (Anthropic, OpenAI, Groq, Mistral, Llama variants) and most serious vector databases (Weaviate, Qdrant, Milvus, Pinecone, Astra DB).

It sits in the same visual-agent-builder niche as Flowise and n8n's AI nodes, but leans harder into the Python developer workflow: any node can be edited as code, and flows compile down to something you can actually run in production rather than a black-box SaaS graph. The core project is open source (MIT, six-figure GitHub stars) and self-hostable; DataStax also offers a hosted Langflow with a free tier plus paid enterprise plans for teams that don't want to run it themselves.

Best thought of as a prototyping surface for people who already know LangChain concepts and want to skip the boilerplate. Non-developers will hit the limits of the abstraction quickly, and the fast-moving codebase means flows built on older versions sometimes need rework after upgrades.