Free forever Windows

Build AI pipelines
that check their own work.

Multiple models collaborate in a visual pipeline — generating, critiquing, and refining each other's output. Design on a canvas or describe what you need in plain text.

Download for Windows How it works

No account Bring your own API key Zero telemetry Offline-capable

Output quality

The difference is visible.

Same prompt, same model. One pass versus a pipeline with a dedicated critique layer and a refinement pass before delivery.

Prompt Design a modern sneaker concept that could be popular in 2030, clean product render, studio lighting. gemini-3.1-flash-image-preview
⨯ Single model Plain prompt, direct generation
Output from a single model without a pipeline
✓ FlowGraph pipeline Prompt shaped by multi-role pipeline
Output produced by a FlowGraph multi-model pipeline

The core problem

One prompt hopes for the best.
A pipeline makes sure.

A single model generates and judges itself in a single pass. A pipeline separates the roles.

Without a pipeline

One model, zero accountability

  • You write a long prompt hoping one model covers every angle
  • The model generates and self-approves in a single pass
  • Output quality depends on your prompt engineering skill
  • Iteration means rewriting the prompt and starting from scratch
  • No way to see how the answer was actually constructed
With FlowGraph Pro

Separate roles, built-in quality gates

  • Dedicated nodes for generation, critique, synthesis, and scoring
  • Generators never judge their own output — an independent Examiner does
  • Design the Flow once, reuse it across any task
  • Cycle mode re-runs automatically until quality hits your target score
  • Every step is visible in the graph — nothing is a black box

The workspace

Three panels. Full visibility.

1

AI Providers

Left sidebar. Add any provider, paste an API key — models appear immediately.

2

Pipeline Graph

Center canvas. Your Flow as a visual graph — nodes connected in layers. Draw edges, rearrange freely, or build from a DNA string.

3

Task & Results

Right panel. Type your prompt, press Run, review the result with Markdown and SVG preview. Enable Examiner for quality scoring (0–100).

FlowGraph Pro
FlowGraph Pro — visual pipeline editor interface

How it works

Four steps to verified output.

Load a Flow, write a brief, collect refined output. The pipeline handles coordination — you focus on the task.

01

Pick or build a Flow

Start from a pre-built Flow, build your own on a blank canvas, or describe what you need in plain text — the app generates the graph for you.

02

Write your brief and run

Enter your brief in the right panel. Assign models to nodes. Press Run.

03

Review with quality scoring

Review results with Markdown and SVG preview. Enable Examiner for quality scoring. Use Cycle mode to re-run automatically until the score meets your target.

04

Export or share

Results save automatically. Export the workspace as a file, or share a Flow with a single DNA string — anyone can load it instantly.

Features

Built for people who don't trust AI blindly.

Pre-built Flows

Six ready-to-use pipelines are included. Load one, enter your brief, run.

Visual Flow Editor

Build pipelines by connecting nodes on a canvas. Each node is a direct LLM call with its own system prompt. Draw edges, reorder layers, adjust prompts inline.

Examiner & Cycle Mode

An independent model scores output 0–100 with a verdict and critique. Cycle mode re-runs the pipeline automatically until the target score is met.

Flow DNA

Every pipeline encodes into a single text string. Copy it, share it, paste it — the full graph rebuilds instantly.

Auto-save & export

Every run, every result, every Flow saves locally. Export the entire workspace as one file to back up or move to another machine.

Mix providers in one pipeline

Ollama, LM Studio, DigitalOcean, Novita AI, or any OpenAI-compatible endpoint. Different models on different nodes — within the same pipeline.

Fully offline with Ollama

Run multi-model pipelines on local hardware. No keys, no cloud, no latency.

Parallel execution & token tracking

Nodes in the same layer run simultaneously. A global counter tracks token usage across the session.

Private by default

No accounts. No analytics. No telemetry. Zero network requests unless you trigger an API call.

Providers

Bring your own models.

Cloud or local. Mix them freely — each node in your pipeline can use a different provider and model.

Ollama Local

Run Llama, Mistral, Qwen, DeepSeek and dozens of other models entirely on your machine. No key, no internet required.

LM Studio Local

Desktop runtime for local models with a built-in OpenAI-compatible server. Great if you prefer a GUI for model management.

DigitalOcean Cloud

Serverless GPU inference from a trusted cloud provider. Good for teams already on the DigitalOcean ecosystem.

Novita AI Cloud

Fast, cost-efficient inference on open-source models. Competitive pricing for high-volume pipeline runs.

Any OpenAI-compatible API Custom

If it speaks the OpenAI chat completions protocol, it works. Point to any endpoint — self-hosted, corporate proxy, or new provider.

Pre-built Flows

Real pipelines, not templates.

Six Flows ship with the app. Each is a working multi-model pipeline — not a placeholder. Write your task, press Run.

Brand

Brand Genesis 360°

Describe a business idea → get a complete brand platform: positioning, messaging, visual direction, risk assessment, and go-to-market strategy.

Pitch

Pitch Deck Architect

Describe your startup → get a slide-by-slide investor narrative with market sizing, unit economics, competitive landscape, and objection stress-testing.

Design

SVG Logo Architect

Describe the brand → get multiple logo variants in SVG code: minimalist, emblem, abstract, typographic. Each variant is critiqued and refined before delivery.

Visual

Aesthetics Architect

Describe the concept → get a visual identity system grounded in color psychology, historical context, and current design trends.

Research

DeepDive Analyzer

Ask a complex question → get a structured synthesis. The pipeline breaks it into sub-queries, runs parallel research tracks, then consolidates and fact-checks.

Creative

Prompt Forge

Describe an image idea → get a polished, production-ready prompt for Midjourney, DALL·E, or Stable Diffusion.

Quick start

Up and running in two minutes.

01

Download & Launch

Run FlowGraphPro.exe. No installation, no admin rights. Fully portable — runs from any folder, any drive.

Self-contained — no runtime install needed.

02

Connect a Provider

Expand a provider in the left sidebar. Paste your API key. Available models load immediately — no restart, no config file.

For offline use, start Ollama locally — no key needed.

03

Open a Flow

Choose a pre-built Flow from the list, or start fresh. Add nodes, connect layers, set system prompts for each. Or paste a Flow DNA string — the graph builds itself.

04

Run & Evaluate

Enter your task in the right panel. Press ▶ Run. Enable Examiner for quality scoring, Cycle for automatic re-runs until your target score is hit.

Flow

A complete pipeline — its graph, node prompts, and settings. You can have multiple Flows open at once, each in its own tab.

Model node

A single LLM call. Has its own system prompt and model assignment. The basic building block of every Flow.

Layer

A horizontal step in the graph. All nodes in a layer execute in parallel, then pass their output to the next layer.

Examiner

An independent model that scores the final output 0–100 with a verdict and critique. Works as a quality gate for Cycle mode.

Flow DNA

A compact text string that encodes an entire Flow. Copy it, share it, paste it — the full graph rebuilds instantly on the other end.

Cycle mode

Automatic iteration. The pipeline re-runs based on the Examiner's feedback, adjusting until the quality score meets your target threshold.

Download

Try it now. Free forever.

No account, no installer. Download, unzip, run. Pre-built Flows are ready on first launch.

Download v1.0.1 — Windows

Windows 10+ · Self-contained — no runtime required · 71 MB portable ZIP

FlowGraph Pro makes zero network requests of its own. The only traffic is the API calls you explicitly trigger to whichever provider you've configured. No analytics, no crash reporting, no telemetry. Verify with Wireshark, Fiddler, or Windows Firewall logs.

Windows SmartScreen may flag the executable as from an unknown publisher — click "More info → Run anyway". You can scan the file with VirusTotal or any antivirus before running.

Privacy Policy Terms of Service License (EULA)