Langflow
Open-source visual builder from IBM for assembling AI agents and RAG applications by drag and drop and serving them as an API.
What it does
Langflow is a visual editor in which you assemble AI workflows from building blocks: language models from various providers, vector databases, document loaders, tools and agents. You can adapt any block in Python or write your own. The Playground lets you test a flow immediately; finished flows are called from your application via an API. Langflow works as an MCP client and can also expose its own flows as an MCP server. You can install it as a desktop app for macOS and Windows, via Docker or as a Python package.
Who it suits
Developers and technically skilled teams who want to prototype chatbots, document search or agents quickly and later run them on their own infrastructure.
Data protection for businesses
Langflow runs on your own computers or servers, so you decide where data is stored – including in the EU. Inputs go only to the model providers you use in a flow, and their training terms apply as well. The software sends anonymous usage data but no prompts, messages or credentials; setting the variable DO_NOT_TRACK=True switches this off. The former hosted version at DataStax was shut down in April 2026. The vendor does not mention a data processing agreement for Langflow. For businesses, IBM offers paid support and integration with watsonx Orchestrate.
Limits
Without Python and server skills you will hardly get beyond simple prototypes. Operation, updates and security are your responsibility. To download the desktop app, IBM asks for your name, email address and company.