How to Use Flowise

Flowise lets you build LangChain flows by dragging nodes on a canvas instead of writing Python.

Flowise is an open-source visual builder for LLM orchestration. You connect nodes — LLMs, prompts, memory, vector stores, tools — on a canvas and it emits a working API endpoint. Under the hood it’s LangChain JS, so anything LangChain does, Flowise can wire up without code.

What it is

A Node.js app with a React Flow canvas. Every node is a thin wrapper around a LangChain class. The runtime executes the graph when a request hits the chatflow endpoint, and the studio lets you test conversations in a side panel while you build.

Install / set up

# npx quickstart
npx flowise start
# or docker
docker run -d -p 3000:3000 \
  -v ~/.flowise:/root/.flowise \
  flowiseai/flowise

First run

Open http://localhost:3000, click New Chatflow, and drop a Chat Model node, a Conversation Chain node, and a Buffer Memory node onto the canvas. Connect them, add your OpenAI key in the model node, and hit the purple chat icon to test.

$ curl -X POST http://localhost:3000/api/v1/prediction/<flowId> \
  -H "Content-Type: application/json" \
  -d '{"question":"hello"}'
{"text":"Hi! How can I help?"}

Everyday workflows

• Start from a marketplace template (RAG, SQL agent, multi-agent) and customize rather than building from scratch.
• Use the Credentials vault so API keys aren’t baked into exported flows.
• Embed the generated chat widget in a website with the provided <script> snippet, no backend glue needed.

Gotchas and tips

Flows are JSON under the hood but node versions change between Flowise releases. Export flows regularly and test after upgrading — a renamed input on a node will silently drop the connection and your agent will misbehave at runtime.

The canvas abstracts LangChain but doesn’t hide its sharp edges. Token windows, streaming quirks, and tool-call loops still bite you. Keep the browser devtools open on the server logs tab while iterating — that’s where the real errors surface.

Who it’s for

Builders who think visually, prototypers who want to demo an agent before committing to code, and teams pairing a non-engineer PM with a LangChain-fluent dev. Production-grade workloads often graduate to raw code, but Flowise gets you to a working prototype fast.