Langflow

Langflow is an open-source, low-code platform for building AI agents and retrieval-augmented generation (RAG) workflows through a visual drag-and-drop canvas. Instead of writing framework code, developers wire together components — models, prompts, tools, vector stores, memory, control flow, and custom logic — into a flow, then run and iterate on it in a built-in playground. It is Python-based and deliberately agnostic to any model, API, or database, so a flow can mix OpenAI, Anthropic, local models, and different vector stores without rewrites.

The visual layer is the core differentiator from code-first agent frameworks: it lowers the barrier for prototyping and lets less code-heavy team members participate, while every flow remains a real, inspectable pipeline rather than a black box. Flows can be exported and served as an API endpoint or as Python code, which is the bridge from a drag-and-drop prototype to something a production application calls. Custom components can be written in Python when built-ins are not enough, so the canvas does not become a ceiling.

Langflow began as an independent open-source project and is now maintained under DataStax (part of IBM), with tight integration to the Astra vector database. The self-hosted version is free under an open-source license and can run locally or on your own infrastructure; a managed cloud offering adds hosted execution with a free entry tier and paid plans for teams needing dedicated infrastructure, higher execution volume, and support. The enormous GitHub following makes it one of the most visible tools in the visual-agent-builder space.