Flowise

Flowise is a visual builder for large language model applications and AI agents. It provides a drag-and-drop canvas, called AgentFlow, for wiring together LLM chains, retrieval-augmented generation pipelines, tools, and multi-agent workflows without writing the underlying orchestration code by hand. The project is organized as a monorepo with separate server, UI, and components packages, and it supports Node.js 20 and above. Documentation and the README are also translated into Traditional Chinese, Simplified Chinese, Japanese, and Korean.

Key features

• Drag-and-drop agent builder: construct chains and multi-agent flows visually on the AgentFlow canvas instead of writing orchestration code by hand.
• LangChain-based node library: pre-built nodes for LLMs, vector stores, memory, and tools built on top of LangChain, exposed through the components package.
• Modular monorepo structure: server handles the API, ui is the React frontend, components holds third-party integrations, and api-documentation generates Swagger docs, each independently maintainable.
• Auto-generated API documentation: the api-documentation package produces Swagger UI docs directly from the Express server's routes.
• Multiple install paths: run via a global npm package, a locally built Docker image, or Docker Compose, depending on how much infrastructure control is needed.
• One-click cloud templates: deployment templates for Railway, Render, Northflank, HuggingFace Spaces, Elestio, Sealos, and RepoCloud, plus documented guides for AWS, Azure, GCP, Digital Ocean, and Alibaba Cloud.
• Active discussion channel: a GitHub Discussions board for questions and feature requests, alongside a Discord community for real-time support.

Getting started

Installing Flowise locally requires Node.js 20 or newer. npm install -g flowise installs the CLI, and npx flowise start launches the app on http://localhost:3000. For a source build, the project uses PNPM as its package manager: clone the repository, run pnpm install to install every module's dependencies, then pnpm build followed by pnpm start. A pnpm dev mode hot-reloads code changes automatically and serves the app on port 8080 once VITE_PORT and PORT are set in the ui and server .env files respectively.

Tech stack

The server package is a Node.js backend, the UI is built with React, and the components package wraps third-party integrations as LangChain-compatible nodes. Contributors needing a larger heap during build can set NODE_OPTIONS=--max-old-space-size=4096 to avoid out-of-memory errors (exit code 134) on pnpm build. Environment variables for the server, documented in the contributing guide, control things like database connection and API keys used by the node integrations. Flowise fits teams that want to prototype or ship an LLM-backed chatbot, internal assistant, or agentic workflow without hand-rolling chain logic for every change, since the AgentFlow canvas lets non-engineers adjust prompts, tools, and routing after an engineer wires up the initial flow. It also works as a shared component library across a team, since integrations built as LangChain nodes in the components package are reusable across any flow built on the canvas, and the same server instance can host multiple independently configured flows.

Flowise is released under the Apache License Version 2.0.