Langflow

Langflow is an open source platform for building and deploying AI-powered agents and workflows through a visual, drag-and-drop interface. It targets developers who want to compose LLM chains, RAG pipelines, and multi-agent systems without hand-writing all the orchestration code, while still allowing full customization through Python. Each flow can be exposed as a REST API or an MCP server, turning a visual pipeline into a tool other applications can call. The project ships both as a Python package for local development and as a bundled desktop app for people who don't want to manage Python environments at all.

Key features

• Visual builder interface: a drag-and-drop canvas, built on React Flow, for composing LLM chains, agents, and RAG pipelines without starting from a blank script.
• Source code access: every component is editable Python, so behavior can be customized beyond what the visual builder exposes out of the box.
• Interactive playground: test and refine a flow step by step, inspecting intermediate outputs before deploying it.
• Multi-agent orchestration: coordinate multiple agents with conversation management and retrieval built in, rather than wiring that logic by hand.
• Deploy as an API or MCP server: export a flow as JSON, call it through a REST API, or expose it as an MCP server for other MCP clients to use as a tool.
• Broad model and vector store support: works with all major LLM providers alongside a growing library of vector databases and AI tools.
• Observability integrations: connect LangSmith, LangFuse, and similar tools to trace and debug flows once they're running in production.

Ideal use cases

Langflow fits teams prototyping retrieval-augmented generation systems, multi-agent workflows, or chatbots that need to move from a whiteboard idea to a running API quickly. Because flows export as JSON or deploy behind an API, they work as building blocks inside larger Python applications rather than as a standalone product. Enterprises evaluating agent orchestration tools use Langflow's visual layer to get non-Python stakeholders involved in the design process while keeping the underlying execution logic in code, and the MCP server option makes a flow reusable as a tool across any MCP-compatible client.

Self-hosting and deployment

Langflow installs as a Python package with uv pip install langflow (Python 3.10 through 3.14 required) and starts with uv run langflow run, which runs a local server on port 7860 by default. A Docker image, langflowai/langflow:latest, is available for containerized deployments on port 7860, and the project documents guides for deploying to major cloud providers. Contributors who clone the repository can run make run_cli from the project root instead of installing the package. A desktop build for Windows and macOS, Langflow Desktop, bundles all dependencies, so users who don't want to manage a Python environment can install it directly from langflow.org.

Tech stack

Langflow's backend and component system are written in Python, and the visual canvas is built with React Flow. The project is licensed under MIT, tracks release notes and download stats publicly on PyPI, and maintains an active Discord community alongside its GitHub repository.