Langflow is a visual builder for AI agents and workflows. You compose flows from components on a canvas, test them in an interactive playground with step-by-step control, and then expose each flow through the built-in API server or as an MCP server so it becomes a tool for applications on any stack. Every component's Python source is editable, so a flow can be customized beyond what the visual editor offers.
It runs as a Python package (Python 3.10 to 3.14), a Docker image, or Langflow Desktop for Windows and macOS, which bundles all dependencies. Flows can also be exported as JSON for use in Python applications, and deployment guides cover the major clouds.
Features
- Visual builder: drag-and-drop composition of flows with iteration on the canvas
- Source code access: any component can be modified in Python
- Interactive playground: run and refine flows step by step before deployment
- Multi-agent orchestration: conversation management and retrieval across agents
- API deployment: each flow is served through a built-in API or exported as JSON
- MCP server: flows become tools for MCP clients
- Observability: integrations with LangSmith, LangFuse, and others
- Model and store support: the major LLMs and vector databases plus a library of AI tools
