How to Install and Use Munder Difflin: A Beginner's Guide

If you have never used an AI coding tool, most guides start three steps ahead of you. This one does not. It starts with what these things actually are, then walks the whole way to a working setup.

You will need about twenty minutes.

First, what is Munder Difflin? #

Munder Difflin is a desktop app that runs a small office of AI workers on your computer.

You see an actual office floor on screen, with characters at desks. Each character is an AI agent working on something you asked for. One of them is your clone, the boss of the floor, who takes what you ask for, breaks it into jobs, and hands those jobs to the others. You talk to your clone, and your clone manages everyone else.

The app is free and open source. It runs on your machine, and it keeps its files on your machine.

The thing it is not: a chatbot. You are not sitting there typing and waiting for replies. You give your clone a job, close the laptop, and come back to work that was done while you were gone.

What is a coding agent, and how can it touch my computer? #

This is the part worth reading slowly, because it is the part that surprises people.

A coding agent is an AI that does not just talk back. It can act. Give a normal chatbot a question and it returns text. Give a coding agent a job and it will read your files, write new ones, run commands, install things, and check its own work, in a loop, until the job is done or it gets stuck.

It does this through a program on your computer called a CLI agent. CLI stands for command line interface, which is the text-only way of controlling a computer that predates windows and buttons. Claude Code, Codex, Antigravity and the others in this guide are all CLI agents. They are the engine. Munder Difflin is the office that runs a team of them and keeps track of everything around them: memory, tasks, schedules, files, and when to come ask you something.

So how does it reach your computer? When you install one of these engines, it runs as a program with your permissions. Anything you can do from your own account, it can technically do: open your files, edit them, delete them, run software, reach the internet. It is not sandboxed away from your machine by default. That is exactly why it is useful, and exactly why you should be deliberate.

Three things keep this sane, and Munder Difflin sets all three up during onboarding:

1. You choose which folders it can work in. Agents work inside the projects you add. A project is just a folder. Do not point your first one at your entire home directory.
2. You choose how much freedom it has. During setup you pick whether agents work on their own or pause and ask you before changing files and running commands. If you are new, pick ask.
3. You can watch it happen. Every agent has a terminal you can open and read, live. Nothing happens off-screen.

Step 1: Which AI engine will you use? #

Munder Difflin does not include an AI. It drives one you install. Twelve engines are supported, but you only need one, and the right one is usually decided by what you already pay for.

Find yourself below.

You do not want to pay anything #

Use Antigravity. It is Google’s agentic development platform and it is available at no charge. It ships with a command line tool called agy, which is what Munder Difflin drives. This is the best starting point for most people reading this guide.

OpenCode is the other free path. It is open source, and it lets you bring your own model, including free and local ones.

You already pay for ChatGPT #

Use Codex. Your ChatGPT Plus or Pro subscription covers it. Nothing more to buy.

You already pay for Claude #

Use Claude Code. Your Claude Pro or Max subscription covers it. This is the engine the app was built against first, and it is the most thoroughly tested of the twelve.

The other engines #

Any of these work. Pick one only if you already use it.

You can add more engines later and run several at once, with different agents on different engines.

Step 2: Install your engine, and log in #

This is the only step where you open a terminal. Once it is done you will not need it again.

The terminal is a plain text window for typing commands. On macOS press Cmd+Space, type Terminal, press Enter. On Windows press the Start button, type PowerShell, press Enter. On Linux you already know.

Find your engine below, copy the block, paste it, press Enter. Wait for it to finish before typing the next line.

First: do you have Node.js? #

Most of these engines install through npm, which comes with Node.js. Check with:

node --version

If that prints a version number, you are ready. If it says “command not found”, get the LTS build from nodejs.org/en/download, install it, then close the terminal window and open a fresh one.

Antigravity, Claude Code and Cursor can be installed without Node.js. Everything else needs it.

Antigravity (free, no subscription) #

Google’s coding agent. The command is agy. Homepage: antigravity.google · Install docs: antigravity.google/docs/cli/install

macOS and Linux:

curl -fsSL https://antigravity.google/cli/install.sh | bash

Windows PowerShell:

irm https://antigravity.google/cli/install.ps1 | iex

Then sign in:

agy

Your browser opens. Sign in with a Google account.

OpenCode (free, bring your own model) #

Open source, works with free and local models. Docs: opencode.ai/docs

npm install -g opencode-ai@latest
opencode auth login

Codex (with a ChatGPT subscription) #

Covered by ChatGPT Plus or Pro. Docs: github.com/openai/codex

npm install -g @openai/codex
codex

On first run it opens a browser to sign you into your ChatGPT account.

Claude Code (with a Claude subscription) #

Covered by Claude Pro or Max. Docs: docs.claude.com/en/docs/claude-code

npm install -g @anthropic-ai/claude-code
claude

If you would rather not install Node.js, use the standalone installer instead.

macOS and Linux:

curl -fsSL https://claude.ai/install.sh | bash

Windows PowerShell:

irm https://claude.ai/install.ps1 | iex

Then run claude and sign in when the browser opens.

The other engines #

Same idea: install, then run the command once on its own to sign in.

On Windows, the Cursor line is irm 'https://cursor.com/install?win32=true' | iex in PowerShell.

Check it actually installed #

Run this, swapping claude for whichever command you installed:

claude --version

A version number means you are done with the terminal. “Command not found” means the install did not finish, or the terminal is still running from before the install: close the window, open a new one, and try again.

Step 3: Install Munder Difflin #

Go to munderdiffl.in and download the build for your system, or take it straight from the latest release on GitHub.

macOS. Download the .dmg. Open it, drag Munder Difflin into Applications. The first time you open it, macOS may say it cannot verify the developer. Right-click the app icon, choose Open, then click Open in the dialog. You only do this once.

Windows. Download the .exe installer and run it. Windows SmartScreen will likely warn you, because the installer is not yet EV code signed. Click More info, then Run anyway. If the installer fails to start at all, download the portable build instead, which needs no installation.

Linux. Download the AppImage, make it executable (chmod +x on the file) and run it. There is a .deb if you prefer to install it properly.

Step 4: Onboarding, six screens #

Open the app. It walks you through setup. There are two intro screens and then four numbered steps.

Screen 1: Who are you? #

It asks whether you are technical or not.

Pick the non-technical option. This is not a downgrade. It changes the language of the entire rest of the app, replacing engine ids, flags and model slugs with plain descriptions. You can change it later in Settings.

Screen 2: Meet your office #

An explanation of what you are about to set up. Read it and continue.

Step 1 of 4: A home for the app #

The app needs an empty folder to keep its own files: its settings, and your agents’ memory.

Something like HarnessAgents in your home folder is fine. It will create it if it does not exist. This is not where your work goes. It is the app’s own filing cabinet.

Step 2 of 4: Your clone #

Name your clone and pick the engine that powers it. This is where the engine you installed in Step 2 shows up. If it is not listed as installed, the sign-in did not finish, so go back and run the command once more.

Give your clone the most capable model available to you. It is doing the thinking and delegating, so it benefits from the extra capability more than the workers do.

Step 3 of 4: Your projects #

Add the folders you want agents working in. A project is just a folder. It can hold code, documents, notes, or anything else.

You can skip this and add projects later. If you are adding one now, make it a new empty folder.

Step 4 of 4: Permissions #

The important screen. It asks how much agents can do on their own.

On means agents carry out tasks without stopping to ask. Smoothest, and how you will eventually want it.

Off means agents pause and ask you before changing files or running commands.

Start with off. Watch a few tasks run, see what the agents actually try to do, and turn it on when you are comfortable. You can change this any time.

Step 5: Your first job #

You are looking at the office floor. Your clone is there. Nobody else yet.

Talk to your clone the way you would talk to a capable new hire who has not met you yet. Be specific about the outcome, and do not bother specifying the steps.

Weak: “organise my files”

Better: “Look in my Documents/receipts folder. Sort every PDF into subfolders by year, based on the date inside the document, not the filename. Tell me about anything you cannot read.”

Your clone will decide it needs a worker, hire one, and hand the job over. You will see a new character appear at a desk.

Click any agent to open its terminal and watch it work in real time.

Things people actually use this for #

Not all of these are coding.

• Documents. Sort a folder of files, rename them consistently, pull the numbers out of fifty invoices into one spreadsheet.
• Research. Give an agent a question and a place to write, and let it work through it while you do something else.
• Code. The obvious one. Fix a bug, add a feature, upgrade a dependency, write the tests.
• Repeating chores. Anything you do every Monday, handed to a schedule instead.
• Long jobs. The ones you keep not starting because they take four hours. Start one before bed.

Step 6: Add agents yourself #

Your clone hires people on its own, but you can hire directly too.

Click add agent. You pick a name, an engine, a model and a project. Different agents can run on different engines, so you can put a free engine on the boring jobs and your paid one on the hard ones.

Agents persist. Close an agent and its memory and inbox reattach when you bring it back, with the same identity.

Step 7: Work that starts without you #

Open the triggers tab. Everything here starts work when you are not typing.

Schedules run a prompt on a repeating clock. You give it a label, choose who it goes to, and write the prompt, which is sent word for word on every run. This is the feature that turns the app from a tool you use into something that runs on its own.

Good first schedule: every morning, ask an agent to check something and write you a short summary.

Context handles what happens as an agent’s memory fills up (see below).

Webhooks let an outside system post work in.

Organisation lets a teammate’s Munder Difflin send work to yours.

Step 8: Context and auto-compaction, briefly #

Every AI model has a limited working memory, called its context. A long-running agent will fill it. When it is full, the agent gets confused or stops.

Auto-compaction is the fix. When an agent’s context gets close to full, the app has it summarise what has happened so far and carry the summary forward instead of the raw history. The agent keeps going without losing the thread.

You control this under Context in the triggers tab: compact automatically, clear entirely, or leave it alone. The default is sensible and most people never touch it.

Step 9: Settings worth knowing #

Seven sections. These are the ones that matter early:

• General. Change your home folder, switch between technical and plain language, pick your language. As of v0.4.6 the app runs in English, Chinese and Arabic.
• Prerequisites. Checks whether your engines are actually installed and logged in. This is the first place to look when something is not working.
• Agents & Models. Which engines are available and which models each uses.
• Autonomy & Budgets. The permission choice from onboarding, plus spending limits per agent. Worth setting a cap before leaving anything running overnight.
• Connections. Slack, webhooks, MCP and the REST API.
• Voice. Talk to your agents out loud.
• Memory & Knowledge. What your agents remember between sessions.

Step 10: Focus mode #

When you want to stop watching the office and work with one agent, press the fullscreen control on its terminal. The floor disappears and you get a full-width terminal.

Press Esc to come back.

When something goes wrong #

An agent dies the moment it starts. Almost always the engine is installed but not logged in. Open Settings, then Prerequisites. Or run the engine’s command once in a terminal and finish the sign-in.

“Engine not installed” during onboarding, but you installed it. The app looks for the command on your system path. Close the app completely and reopen it, since it reads your environment at launch.

Windows blocks the installer. SmartScreen flags it because the installer is not EV code signed. Choose More info, then Run anyway, or use the portable build.

macOS says the developer cannot be verified. Right-click the app, choose Open, then Open again.

Nothing is happening. Check whether message delivery is paused in the Command Center. Paused means queued work is being held for every agent, and nothing is lost when you switch it back on.

The short version #

1. Pick an engine. Antigravity if you want free, Codex if you pay for ChatGPT, Claude Code if you pay for Claude.
2. Install it and sign in. One terminal command. The only one.
3. Download Munder Difflin and install it.
4. Six onboarding screens. Pick the non-technical path, use a new empty folder, start with agents asking permission.
5. Give your clone a real job and watch what happens.

The app is open source on GitHub. If you get stuck, the Discord is the fastest place to ask.

FAQ

Do I need to know how to code to use Munder Difflin?

No. The onboarding asks whether you are technical or not on its very first screen, and the non-technical path replaces every piece of jargon with plain language. You will open a terminal exactly once, to install and log into your AI engine, and you can copy and paste those commands.

Does Munder Difflin cost anything?

The app is free and open source. What you may pay for is the AI engine behind it. Antigravity is available at no charge and OpenCode has a free path, so you can run the whole thing without paying anything. If you already pay for ChatGPT or Claude, you can use those subscriptions instead.

Can these agents really change files on my computer?

Yes, and that is the point, so it is worth understanding before you start. An agent can read files, write files and run commands in the folders you give it. You choose how much freedom it has during setup, and you can set it to ask permission before every change.

Which operating systems does Munder Difflin run on?

macOS, Windows and Linux. macOS ships as a .dmg, Windows as an installer and a portable build, and Linux as an AppImage and a .deb.

Is my code or data sent anywhere?

The app runs on your machine and stores its files there. Your prompts and the files an agent reads do go to whichever AI engine you picked, the same as if you used that tool directly. Anonymous usage telemetry is opt-out and every event is listed publicly in TELEMETRY.md.

---