Flyde - Visual AI Workflows in Your VS Code

TL;DR

TL;DR: Flyde is an open-source VS Code extension that brings drag-and-drop visual programming to backend AI workflows — design prompt chains and agentic flows without sacrificing the ability to ship real TypeScript code.

What Is Flyde?

Flyde is an open-source visual programming editor for backend logic that runs inside VS Code. Unlike standalone workflow tools (n8n, LangFlow), Flyde graphs compile to plain TypeScript modules — real files in your repo that you debug, version, and deploy like any other code.

The project targets teams building LLM-powered products. Product managers, designers, and backend developers all work on the same visual canvas, while engineers retain full control of the underlying TypeScript.

From the README:

Visual AI Flows. In Your Codebase. Open-source, runs in VS Code. Integrates with existing TypeScript code.

Core features (verified from source):

• VS Code extension — edit flows side-by-side with your existing files
• Runtime library — @flyde/runtime for executing flows in Node.js/bun/Deno
• In-codebase integration — flows live as .flyde.ts files you import normally
• AI agent support — build agentic workflows with tool calling, memory, and prompt chaining
• Visual debugger — step through flow execution node by node
• TypeScript throughout — every flow is typed end-to-end

Setup

Try the Playground

No install needed. Visit https://www.flyde.dev/playground to explore the editor in your browser.

Install the VS Code Extension

code --install-extension Flyde.vscode

Or search for “Flyde” in the VS Code Marketplace and install from there.

Start a New Project

npx create-flyde-app

This scaffolds a new Flyde project with a sample AI flow you can run immediately.

Add to an Existing Project

npm install @flyde/runtime

Import and use flows like any TypeScript module:

import { executeFlow } from "@flyde/runtime";
import { MyFlow } from "./flows/my-flow.flyde";

const result = await executeFlow(MyFlow);

Deeper Analysis

How Flows Compile to TypeScript

A Flyde flow is a .flyde.ts file containing a directed graph of nodes (functions or external integrations) and connections (data pipes). When you run a flow, Flyde compiles it to a plain TypeScript function graph that uses the runtime’s executeFlow API.

This means:

• No proprietary runtime to deploy — the compiled output is just TypeScript
• You can commit .flyde.ts files to git alongside your regular code
• Refactoring tools (Go to Definition, Find All References) work across flow boundaries

AI Agentic Workflows

Flyde ships with dedicated nodes for building LLM-based agents:

• LLM Node — configurable prompt, model (OpenAI, Anthropic, local), and temperature
• Tool Node — define functions the agent can call at runtime
• Memory Node — conversation history and state persistence
• Branch Node — conditional routing based on agent output

You wire these together visually to prototype agent behavior, then swap implementations behind the typed interface.

Standard Library

The @flyde/stdlib package includes 40+ pre-built nodes covering common backend tasks: HTTP requests, JSON transforms, date formatting, webhook handlers, and more.

Collaboration

Because flows are code files, the usual PR workflow applies. Flyde also supports a read-only playground shareable link so non-engineers can review flows without needing the extension installed.

Practical Evaluation Checklist

• [x] VS Code extension — install from Marketplace
• [x] Runs locally, no cloud dependency for execution
• [x] Flows compile to TypeScript (verifiable in .flyde.ts files)
• [x] AI agent nodes available (LLM, Tool, Memory, Branch)
• [x] MIT-licensed runtime (@flyde/runtime)
• [x] Discord and GitHub community active

Security Notes

• Flows execute arbitrary TypeScript — treat .flyde.ts files like any trusted code in your repo
• The runtime supports sandboxed execution via Node.js VM module if you need to run untrusted flows
• No telemetry is sent by the runtime by default; check the privacy policy for the VS Code extension

FAQ

Q: How is this different from n8n or LangFlow? A: n8n and LangFlow run as standalone services with their own runtime and database. Flyde flows are TypeScript files that compile into your project — they ship with your app, not as a separate service to operate.

Q: Do I need TypeScript experience to use it? A: Basic familiarity helps since flows are .ts files, but the visual editor abstracts most of the typing. Non-engineers can use the playground and share links without writing code.

Q: Can I use my own LLM provider? A: Yes. The LLM node supports OpenAI, Anthropic, and any OpenAI-compatible API. Local models via Ollama are also supported.

Q: Is there a hosted version? A: No cloud offering exists yet. The runtime is self-hosted and runs wherever your Node.js app runs.

Source and Accuracy Notes

• Project page: flyde.dev
• Source repository: github.com/flydelabs/flyde
• License: MIT for @flyde/runtime; AGPL-3.0 for the flow editor component
• VS Code Marketplace: Flyde.vscode
• HN launch thread: news.ycombinator.com/item?id=44703298
• Stars: 3,500+ (GitHub, June 2026)