ArchHub19 connectors · MIT

Getting started with ArchHub

ArchHub is a desktop app for architects and design studios. It gives you a visual canvas where you connect your real tools — Revit, AutoCAD, Rhino, Excel, Word, Outlook, and more — into a graph you can read, run, and reuse. You drive it by typing what you want into a composer, and it drafts the steps for you to approve.

This guide takes a brand-new user from nothing to a first working result. You do not need to be a developer, and you do not need to touch a terminal.

1. Install ArchHub (Windows, click-only)

There is one supported way to install ArchHub today: the direct download.

  1. Go to github.com/Fargaly/ArchHub/releases/latest.
  2. Download the file named ArchHub-Setup-0.exe.
  3. Double-click it and follow the installer.

Two numbers, one product. The release series is the version you see on the changelog — that is the app’s version. The installer file and its GitHub tag still carry the open-beta packaging label 0, which is why the download is called ArchHub-Setup-0.exe and sits under the v0 tag. They are the same build.

Heads-up about the Windows warning. Windows SmartScreen may show a blue “Windows protected your PC” dialog the first time, because the app is not yet code-signed. Click More info → Run anyway to continue. (Code signing is on the roadmap.)

Package managers are not available yet. scoop, winget and Chocolatey are all unpublished: winget install Fargaly.ArchHub does not resolve, and the scoop manifest in the repository does not match the published installer, so scoop install fails its hash check. Use the direct download above.

The installer creates a desktop icon and a Start-menu shortcut, so after this you launch ArchHub like any normal Windows app.

Running from source (any OS, optional, for technical users). Clone the repository, then run pip install -r app/requirements.txt followed by pythonw app/main.py. Most users should use the installer above.

2. First launch

Double-click the ArchHub icon. A single application window opens — this is the whole product; everything happens inside it.

On the very first run, a short sign-in step appears. ArchHub signs you in to a cloud account so it can give you a chat-capable AI to drive your graphs:

  1. Enter your email address.
  2. ArchHub sends you a magic link — open the email and click the link to confirm. (There is no password to create.)
  3. You land back in the app, signed in on the free open-beta tier. No credit card, no trial clock.

Note on sign-in options. Two paths work today: the email magic link, and Sign in with Google. Pick whichever you prefer — both land you in the same account.

Note on the AI model. A permanently free, zero-setup cloud model is still being finalized, so you may be asked to add your own AI provider key. You can also point ArchHub at a local model you run yourself (see Settings below) — those are detected automatically and preferred when available.

If the window opens but stays blank

On some NVIDIA graphics cards the embedded view can render as a blank window. If that happens, set this Windows environment variable and relaunch:

ARCHHUB_VERIFY_NO_GPU=1

(Open Start → “Edit the system environment variables” → Environment Variables → New, name it ARCHHUB_VERIFY_NO_GPU, value 1.) This switches to software rendering and fixes the blank screen. An automatic fallback is on the roadmap.

3. The Home screen — what you’re looking at

When you’re signed in, you land on Home. Here’s what’s on it:

  • The ArchHub wordmark at the top.
  • Your recent sessions — a grid of cards (with thumbnails) for the graphs you’ve worked on. It’s empty on day one; this is where your work will collect.
  • An account chip showing your real account state — your email, your plan, and how many AI messages you have remaining. (These are live numbers, not placeholders.)
  • A graph-health indicator that, once you have a graph open, tells you whether it’s wired up cleanly. Click it for the detail.

Think of Home as your project shelf: each session is one canvas you can reopen.

4. Your first win — run something on the canvas

Here’s the shortest path to seeing ArchHub actually do work.

Step 1 — Open a canvas

From Home, create a new session. This opens the canvas — the open workspace where your graph lives. You’ll see a toolbar and an empty grid.

Step 2 — Tell the composer what you want

At the bottom of the canvas is the composer — a text box you type into, like a chat. Describe what you want in plain language (for example, “list the sheets in my open Revit model” or “read the values from my open spreadsheet”).

ArchHub turns your request into a plan: a set of steps shown for you to review. Nothing runs automatically without your say-so. You choose how much control you want:

  • Plan mode — ArchHub proposes the steps and waits; you approve each one.
  • Auto mode — ArchHub runs the approved steps for you.
  • YOLO mode — for when you trust it to run straight through.

Approving a step is how a plan becomes a real action. The history of these plans is saved with the session, so you can look back at what ran.

Tip: You can also click the microphone in the composer to dictate instead of type.

Step 3 — Add a tool (connector) node

ArchHub ships with 19 connectors covering 155 operations — the real apps architects use. To add one to your canvas, open the library with Cmd-K (or the library button), search for the app you want, and drop its node onto the canvas.

Each connector operation is honest about its status: it tells you whether the target app is live (running and reachable), loaded but not running, not installed, or needs authorization. So you always know whether a step can actually run before you run it.

A few examples of what’s available out of the box:

  • Revit — list sheets, walls, doors, windows, levels, rooms, families, warnings; set parameters; place tags. (19 operations.)
  • AutoCAD — list layers, blocks, entities, layouts, xrefs; run commands.
  • Rhino — read document info, layers, objects; run scripts.
  • Excel / Word / PowerPoint / Outlook — read and write workbooks, find and replace in documents, build slides, draft email.
  • Photoshop / Illustrator / InDesign — list documents, layers, artboards; export.
  • Procore / Notion / Dropbox / Teams / Speckle — RFIs, submittals, pages, files, messages, model exchange.

The desktop apps (Revit, AutoCAD, 3ds Max, Blender) connect through a small helper that ArchHub installs into the app; Office and Adobe apps connect directly; cloud services connect over the web. The first time you use a desktop connector, ArchHub will tell you what one-time activation step it needs (for example, an add-in for Revit).

Step 4 — Run it and read the result

Run the operation (via your approved plan, or by running the node directly). The result comes back into the canvas, where you can wire it into the next node. That wiring — output of one tool feeding the input of the next — is the whole idea: you build a small, readable pipeline you can run again tomorrow.

That’s your first win: a real operation, against a real tool, on a canvas you can save and reopen.

5. What else is in the app

Once you’re comfortable, these are the other surfaces worth exploring:

  • Library — browse and search every node type, create your own node types, and save a useful arrangement as a reusable skill.
  • Sessions — create, open, rename, duplicate, fork, or delete your graphs, and sync them to the cloud so they follow you between machines.
  • The Brain — a personal memory that learns from your work. You can browse it as folders, search your facts, and back it up to the cloud. (Your brain backup is write-only and has secrets stripped before it leaves your machine.)
  • Settings — choose your AI providers and see router status, set permissions, pick a theme, and adjust accessibility preferences. This is also where you point ArchHub at a local model if you run one. Account details (email, plan, messages remaining) open here too.
  • Self-Heal Inspector — shows when ArchHub has repaired its own wiring.

What’s still coming. ArchHub is in open beta. A couple of things are built but not yet switched on for everyone: a self-extension loop that lets the agent build new connectors for you (today it only proposes; it never applies on its own), and a permanently free cloud model. The features described above are live today, Google sign-in included.

6. Using ArchHub from the website

You can also start from archhub.io:

  • Download takes you to the desktop installer.
  • Sign in / Get started runs the same email magic-link flow as the app.
  • Account shows your plan and how many messages you have remaining.
  • Features, Pricing, Security, Community, and Changelog pages explain the product in more depth.

The desktop app is where the real work happens; the website is for downloading, signing in, and managing your account.

You’re set

You’ve installed ArchHub, signed in on the free tier, opened a canvas, and run a real operation against one of your tools. From here, keep stacking nodes into graphs, save the good ones as skills, and let the Brain learn your patterns.

Welcome to ArchHub — drafted, not generated.