Skip to main content
All products
AI Flow

AI Flow

Visual builder for AI workflows

Drag-and-drop canvas for orchestrating LLMs, tools, and human review. Versioned, observable, and deployable as an API in minutes.

GWEB · AI

InputLLMToolAPIRAGEvalOutputLog
Running · 63%
ai-flow.gweb
What you get

Everything you need to ship AI Flow.

01

Visual DAG editor

Drag-and-drop canvas for orchestrating LLMs, tools, and human review. Versioned, observable, and deployable as an API in minutes.

Visual DAG editor

02

Built-in evals & guardrails

Drag-and-drop canvas for orchestrating LLMs, tools, and human review. Versioned, observable, and deployable as an API in minutes.

Built-in evals & guardrails

03

One-click API deploy

Drag-and-drop canvas for orchestrating LLMs, tools, and human review. Versioned, observable, and deployable as an API in minutes.

One-click API deploy

04

Hybrid human-in-the-loop

Drag-and-drop canvas for orchestrating LLMs, tools, and human review. Versioned, observable, and deployable as an API in minutes.

Hybrid human-in-the-loop
Live demo

Try AI Flow in your browser.

A real, working sample — no signup, no install. Click around and see what shipping with AI Flow feels like.

ai-flow · sandbox Live
IntakeClassifyDraftReviewSend

Built on GWEB · Works with

Drop-in integrations with the tools your team already runs.

OaOpenAIAnAnthropicPcPineconeSnSnowflake
Our AI triage workflow that used to take a team of three to maintain now ships itself. We deploy 4× a day.

Marco Lindgren

Head of Support Ops · Housing MCP

AI Flow impact

4× deploys / day

Get started

Live with AI Flow in a day.

Step 01

Open the canvas

Drag a model, a tool, and a router.

Step 02

Wire it up

Connect, eval, and version in one click.

Step 03

Deploy as an API

Ship the workflow behind a public endpoint.

85k

Workflows shipped

11 min

Avg time to deploy

Start building with AI Flow.

Sandbox keys in 60 seconds. Move to production when you're ready.

gweb · ai-flow
import { GwebAI } from "@gweb/sdk";

const ai = new GwebAI({ apiKey: process.env.GWEB_KEY });

const reply = await ai.ai.flow.invoke({
  input: "Hello, world",
  stream: true,
});