Langflow
Wire up AI agents and RAG pipelines on a visual canvas, then export the finished flow as a working API or MCP server. Free and self-hosted since DataStax's cloud tier closed.
Overview
Langflow puts a drag-and-drop canvas in front of agent and RAG development, connecting large language models, vector stores, memory, and tools into a working flow without hand-written orchestration code. Finished flows export directly as a REST API or a Model Context Protocol (MCP) server, so a prototype built visually can be dropped straight into another application. DataStax's hosted cloud version shut down in April 2026, so Langflow is now exclusively self-hosted, available via Docker, pip install, or a downloadable desktop app for local development. With more than 150,000 GitHub stars and a large, active contributor base, it remains one of the most widely used open-source tools for visually assembling agents and retrieval pipelines. Because the software itself carries no license fee, the only ongoing costs are your own infrastructure and whatever LLM or vector database usage the flow consumes. Langflow fits developers and technical teams who want to prototype or ship agent workflows visually while keeping full control over where the flow runs.
Key Features
- Drag-and-drop visual canvas
- Export flows as an API or MCP server
- Built-in RAG and vector store components
- Desktop app plus Docker/pip self-hosting
- • Fully free since the hosted cloud tier was discontinued
- • MIT licensed with an active, large contributor base
- • Exports directly to a usable API or MCP server
- • No first-party managed hosting anymore, self-hosting is the only option
- • Production usage still costs LLM API tokens and infrastructure