Move your Langflow flows to LiveGraph

Langflow already thinks in nodes and edges — the move is shorter than it looks, and the canvas on the other side stays live while a run executes.

How the move works

There's no Langflow file importer — our importer covers Flowise exports only, and we'd rather say that than half-convert your flow. The honest path is generate, not import:

• Export the flow from Langflow — it downloads as JSON, and a flow file is just the nodes, edges, and component config you already see on its canvas.
• Paste it into the describe box when you create a graph here — or describe the flow in words if the JSON is unwieldy.
• Review the generated graph. It lands as an ordinary editable LiveGraph graph — rename nodes, tighten prompts, rewire edges, then run it.

What a flow maps to

What carries over

• The flow shape. Which component feeds which — that's already a graph, so it regenerates faithfully from the export.
• Prompts. Every instruction and system message inside your components moves into node prompts verbatim.
• Model choices. Each node picks its own provider and model — Anthropic, Google, Groq, OpenRouter, or an OpenAI-compatible endpoint.

What doesn't move

• Credentials and API keys. Provider keys and tool connections in Langflow stay there — each node's tools are re-entered here, encrypted at rest and scoped to that node.
• Custom Python components. Component code doesn't execute on the canvas — rewire the same capability as an MCP server or HTTP endpoint the node can call.
• Vector store connections. The store and its embeddings config don't port; retrieval goes through a tool the node calls.
• The Playground. LiveGraph orchestrates agents against your systems — it doesn't host an embedded chat surface for end users.

Why move

In Langflow you test a flow in the Playground and read the log when it finishes — the canvas is a build surface, not a live one. On LiveGraph the run executes on the same canvas you built: you watch each hop as it happens, a run parks at an approval gate until a person says yes, and dragging an edge mid-run steers the run you're watching instead of restarting it. Generated graphs and hand-built ones get the same treatment — nothing lands in a black box.

Start

Create an account, paste your simplest flow's export into the describe box, and run it once — or watch the live demo first: a real multi-agent graph running in your browser, no signup.

Sign up freeAlready have an account? Log in