Viral Carousels with Gemini CLI

Use Gemini CLI and MCP to generate Instagram carousels in Paper Design.

Introduction

30 Second Summary

Creating Instagram carousels is one of the best ways to grow your audience, but designing each slide from scratch takes forever. What if you could type a single sentence and watch a full set of styled slides appear on a design canvas?

In this project, you will connect Gemini CLI (a free AI agent) to Paper Design (a design tool) using MCP. You will generate a complete 5-slide Instagram carousel and export it as ready-to-post images.

What You'll Build

A repeatable AI-powered workflow where you type a topic into Gemini CLI and it generates a complete 5-slide Instagram carousel inside Paper Design, ready to export and post.

By the end of this project, you'll have:

  • A connected Gemini CLI + Paper Design MCP pipeline that gives a free AI agent direct access to a design canvas.
  • A full set of 5 styled Instagram carousel slides at 1080x1350px, exported as PNG images.
  • A reusable GEMINI.md instruction file that makes future carousel generation a one-liner.
  • Secret Mission: Recreate a carousel from a reference screenshot by feeding an image to Gemini CLI and letting it match the style on your Paper canvas.

Are there any prerequisites?

None. This project includes full setup instructions for every tool you need, including Node.js, Gemini CLI, and Paper Design Desktop. All you need is a Linux computer and a free Google account.

Install Gemini CLI and Node.js

To build your AI-powered carousel system, you need two tools installed on your machine.

The first is Node.js, a JavaScript runtime that lets you run tools from the npm package registry. The second is Gemini CLI, a free AI agent from Google that you control from your terminal.

In this step, get ready to:
  • Install Node.js on your Linux machine.
  • Install Gemini CLI globally using npm.
  • Authenticate Gemini CLI with your Google account.
Check if Node.js is installed

Gemini CLI is distributed as an npm package, so Node.js must be present first.

  • Open your terminal by pressing Ctrl+Alt+T.
  • Check whether Node.js is already installed by running this command:
node --version

✔️ I see a version number

You already have Node.js installed. Move on to the next substep.

ⓧ Command not found

Node.js is not installed yet. Install it by running this command:

sudo apt install nodejs npm -y

This installs both Node.js and npm (the Node Package Manager) from your system's package repository.

  • After the installation finishes, confirm it worked by running:
node --version

You should see a version number printed (e.g. v18.19.0 or similar).

Still seeing command not found?

Try closing and reopening your terminal, then run node --version again. If the issue persists, try installing Node.js via nvm instead.

help me install Node.js on my Linux machine

Install Gemini CLI

Gemini CLI is Google's open-source AI agent for the terminal. It connects to Gemini models using your personal Google account for free.

  • Install Gemini CLI globally by running this command:
npm install -g @google/gemini-cli

What does this command do?

The -g flag installs the package globally, meaning you can run gemini from any directory on your machine.

The package @google/gemini-cli is the official Gemini CLI from Google.

  • Verify the installation succeeded by running:
gemini --version

You should see a version number printed in your terminal.

Seeing a permission error?

If you get an EACCES permission error, your npm global directory might need elevated permissions. Try running the install command with sudo:

sudo npm install -g @google/gemini-cli

help me fix npm global install permissions on Linux

Launch and authenticate Gemini CLI

Gemini CLI needs a one-time authentication with your Google account. This gives you access to the free tier with 1,000 model requests per day.

  • Start a Gemini CLI session by running:
gemini

What happens next?

Gemini CLI will ask you how you want to authenticate. It opens a sign-in page in your web browser where you log in with your Google account. Your credentials are cached locally for future sessions.

  • When Gemini asks how you want to authenticate, select Sign in with Google.
  • A browser tab opens. Select your Google account and click Sign in.
  • Return to your terminal after signing in.

You should now see the Gemini interactive prompt in your terminal, ready to accept messages.

Do I need a paid account?

No. Any personal Google account works. The free tier gives you 1,000 requests per day, which is more than enough for this project.

Browser not opening?

If the browser sign-in page does not open automatically, look for a URL printed in the terminal output. Copy and paste it into your browser manually.

help me authenticate Gemini CLI when the browser does not open

Your tools are ready to go. Next up, you will install Paper Design Desktop and connect it to Gemini CLI via MCP.

Install Paper Design Desktop

Gemini CLI is ready to go. Now you need a design canvas for Gemini to draw on.

Paper Design is a browser-based design tool that also ships as a desktop app. The desktop app is special because it runs a local MCP server in the background. This server is what lets Gemini CLI reach into the canvas and create shapes, text, and artboards directly.

In this step, get ready to:
  • Download and run Paper Design Desktop on Linux.
  • Sign in and create a blank file on the canvas.
  • Confirm the MCP server is running automatically.
Download and run the Paper Design AppImage

Paper Design Desktop is distributed as a Linux AppImage. An AppImage is a single file that works like a portable app. You download it, mark it as executable, and run it.

  • Open your browser and go to paper.design/downloads.
  • Click the download button under Paper Desktop (v0.3.2) to download the Linux AppImage.
  • Open your terminal.
  • Navigate to where the file downloaded (usually your Downloads folder) by running this command:
cd ~/Downloads
  • Make the AppImage executable and run it by running these commands:
# Make the AppImage file executable
chmod +x ./Paper*.AppImage

# Run Paper Design Desktop
./Paper*.AppImage

What do these commands do?

  • chmod +x gives the file permission to run as a program. Linux requires this step for any downloaded executable.
  • ./Paper*.AppImage launches the file. The * wildcard matches whatever version number is in the filename so you don't need to type it exactly.
  • Paper Design Desktop should open in a new window.

What is an AppImage?

An AppImage bundles an entire application into a single file. It runs on most Linux distributions without installing anything to your system. When you close the app, nothing is left behind except the file itself.

Paper Desktop not opening?

  • Make sure the file downloaded completely. If the download was interrupted, re-download it from paper.design/downloads.
  • If you see a "permission denied" error, double check you ran chmod +x on the correct file.
  • If the app opens but sign-in links don't work, you may need to integrate the AppImage with your desktop. Try using Gear Lever, AppImageLauncher, or AppMan to register the protocol handler.

Help me troubleshoot running the Paper Design AppImage on Linux.

Sign in and create a new file

Paper Design needs an account so your files are saved to the cloud. The free tier gives you 100 MCP tool calls per week, which is plenty for this project.

  • When Paper Design opens, click Sign in or Create account.
  • Follow the prompts to create a free account (or sign in if you already have one).
  • Once signed in, click New file to create a blank canvas.
  • You should see an empty canvas with Paper Design's toolbar. Keep this window open for the rest of the project.

Keep Paper Design running

The MCP server only runs while Paper Design has a file open. If you close the app or close the file, Gemini CLI loses its connection to the canvas.

Sign-in not completing?

  • Some Linux desktop environments don't handle the paper:// protocol link automatically. If clicking the sign-in confirmation in your browser doesn't bring you back to Paper, try restarting the app and signing in again.
  • If you're behind a corporate VPN or proxy, you may need to allowlist *.paper.design in your network settings.

Help me fix Paper Design sign-in issues on Linux.

Verify the MCP server is running

Paper Design's MCP server starts automatically whenever a file is open in the desktop app. There is no separate setup needed.

  • Confirm you have Paper Design Desktop open with your new blank file loaded.
  • The MCP server is now running at http://127.0.0.1:29979/mcp. You can optionally verify it by running this command in a new terminal tab:
# Check if Paper's MCP server is responding
curl -s -o /dev/null -w "%{http_code}" http://127.0.0.1:29979/mcp

What should I see?

If the MCP server is running, curl returns an HTTP status code (like 200 or 405). Any response means the server is alive and listening.

If you see 000 or a connection refused error, Paper Desktop is either closed or has no file open. Switch back to Paper and make sure a file is loaded.

MCP server not responding?

  • Make sure Paper Design Desktop is open with a file loaded. The MCP server only starts when a file is open.
  • Try closing and reopening Paper Design Desktop, then create a new file again.
  • If the issue persists, restart your computer and relaunch Paper Design.

Help me get Paper Design's MCP server running on Linux.

Paper Design is running and its MCP server is listening. Next up, you'll connect Gemini CLI to Paper so the AI can actually reach the canvas.

Connect Gemini CLI to Paper Design via MCP

Gemini CLI and Paper Design are both running, but they don't know about each other yet. To make Gemini control your Paper canvas, you need to tell it where Paper's server lives.

MCP (Model Context Protocol) is the bridge that lets AI agents talk to external tools. Paper Design runs a local MCP server whenever you have a file open. In this step, you'll point Gemini CLI at that server so it can create and edit designs on your canvas.

In this step, get ready to:
  • Exit your current Gemini CLI session and create a settings directory.
  • Configure the MCP connection to Paper Design.
  • Verify Gemini CLI connects to Paper's MCP server.
Exit Gemini CLI and create the settings directory

Gemini CLI reads its configuration from a file called settings.json inside a hidden ~/.gemini folder. You need to exit your current session, create that folder, and open the file for editing.

  • In your active Gemini CLI session, exit by typing this command:
/quit

You should see your regular terminal prompt return.

  • Create the Gemini settings directory and open a new settings file by running these commands:
mkdir -p ~/.gemini
nano ~/.gemini/settings.json

What do these commands do?

  • mkdir -p ~/.gemini creates a hidden folder called .gemini in your home directory. The -p flag means it won't show an error if the folder already exists.
  • nano ~/.gemini/settings.json opens a simple terminal text editor where you'll paste the MCP configuration.

You should see nano's text editor open with an empty file. The cursor will be at the top left.

Nano not found?

If you see a "command not found" error, install nano by running sudo apt install nano and then re-run the nano command above.

Help me troubleshoot opening a text editor on Linux

Add the Paper MCP server configuration

This JSON file tells Gemini CLI where to find Paper Design's MCP server. You'll add an entry pointing to the local address that Paper Desktop exposes automatically.

  • Paste the following configuration into nano:
{
  "mcpServers": {
    "paper": {
      "httpUrl": "http://127.0.0.1:29979/mcp"
    }
  }
}

What does this configuration do?

  • mcpServers is the section where you list all the MCP tool servers Gemini should connect to.
  • paper is the name you're giving this server. You can call it anything, but "paper" keeps it clear.
  • httpUrl tells Gemini to connect over HTTP to http://127.0.0.1:29979/mcp. This is the local address Paper Desktop exposes whenever a file is open.
  • Save the file by pressing Ctrl+O, then press Enter to confirm the filename.
  • Exit nano by pressing Ctrl+X.
  • Confirm the file was saved correctly by running:
cat ~/.gemini/settings.json

You should see the JSON configuration printed in your terminal, matching what you just pasted.

Seeing an empty file or no output?

Make sure you pressed Ctrl+O (the letter O, not zero) to write the file before exiting. If the file is empty, re-open it with nano ~/.gemini/settings.json and paste the JSON again.

Help me fix my settings.json file

✔️ Awesome, I've got everything!

Great. Make sure you've saved the file and exited nano before continuing.

ⓧ I'd like to double check the full code

Here is the complete ~/.gemini/settings.json file:

{
  "mcpServers": {
    "paper": {
      "httpUrl": "http://127.0.0.1:29979/mcp"
    }
  }
}
Verify the MCP connection

Now that the configuration is saved, Gemini CLI will read it on startup and attempt to connect to Paper's MCP server. Let's confirm the connection works.

  • Start a new Gemini CLI session by running:
gemini
  • Once you see the Gemini interactive prompt, check the MCP connection by typing:
/mcp

You should see the Paper server listed with a status of CONNECTED and its available tools displayed (like create_artboard, write_html, export, and others).

Paper server showing DISCONNECTED?

This means Gemini can't reach Paper's MCP server. Check these common causes:

  • Make sure Paper Design Desktop is still open with a file loaded. The MCP server only runs when a file is open.
  • Check that the httpUrl value in your settings.json is exactly http://127.0.0.1:29979/mcp with no trailing slash or typos.
  • Try exiting Gemini CLI with /quit and running gemini again. Sometimes the connection needs a fresh start.

Help me troubleshoot the Paper MCP connection

Your AI-to-canvas pipeline is connected. Next up, you'll test it by asking Gemini to create something on your Paper Design canvas.

Test the Connection with a Red Rectangle

Your Gemini CLI is connected to Paper Design via MCP. But a connection status alone does not prove the full pipeline works end to end.

In this step, you will send a simple prompt to Gemini and watch it create a shape directly on your Paper canvas. If the red rectangle appears, you know for certain that AI instructions flow all the way from your terminal to the design tool.

In this step, get ready to:
  • Send a test prompt from Gemini CLI to Paper Design.
  • Confirm a shape appeared on your canvas.
  • Delete the test shape to prepare a clean workspace.
Send a test prompt to Gemini CLI

You will type a natural language instruction into your active Gemini CLI session. Gemini will interpret it and call the Paper MCP tools to create the shape.

  • In your Gemini CLI session, type the following prompt and press Enter:
Create a red rectangle in Paper.

What happens next?

Gemini reads your prompt and decides which MCP tools to call. It will show you a confirmation dialog with the tool name and arguments it wants to execute.

This is a safety feature. Gemini always asks for your permission before calling an MCP tool for the first time.

  • When Gemini shows the tool confirmation prompt, select the option to proceed (you can choose to allow it once, or always allow tools from the Paper server).

What are the confirmation options?

Gemini gives you four choices when an MCP tool call needs approval:

  • Proceed once executes the tool this time only.
  • Always allow this tool auto-approves this specific tool in future calls.
  • Always allow this server auto-approves all tools from the Paper server.
  • Cancel stops the tool call.

Choosing Always allow this server is the most convenient option. It means you will not be prompted again for any Paper tool calls during this session.

Confirm the rectangle appeared on your canvas
  • Switch to Paper Design (click the Paper Design window or use your taskbar).
  • Look at your canvas. You should see a red rectangle that was not there before.

This red rectangle proves the full AI-to-canvas pipeline is working. Gemini successfully called the Paper MCP tools, and Paper executed the instructions on your canvas.

Don't see a red rectangle?

  • Make sure you approved the tool call in Gemini CLI. If you selected Cancel, Gemini will not execute the tool. Switch back to your terminal and try the prompt again.
  • Confirm that Paper Design still has a file open. The MCP server only runs when a file is active in Paper Desktop.
  • Zoom out on your Paper canvas. The rectangle may have been placed outside your current viewport.

Still stuck? Why isn't Gemini CLI creating shapes on my Paper Design canvas?

Delete the test rectangle

Now that you have confirmed the pipeline works, remove the red rectangle so you have a clean canvas for your carousel project.

  • In Paper Design, click on the red rectangle to select it.
  • Press the Delete key (or Backspace) to remove it from the canvas.
  • Confirm your canvas is now empty and ready for the next step.

You can also ask Gemini to clean up

Instead of manually deleting, you could type a prompt like "Delete everything on the canvas" in your Gemini CLI session. Gemini would use the Paper delete_nodes MCP tool to remove the rectangle for you.

Your pipeline is proven and your canvas is clean. Next up, you will use this same workflow to generate a full set of Instagram carousel slides.

Generate Your First Carousel

Your AI-to-canvas pipeline is proven. A red rectangle appeared in Paper Design the moment you asked for it. Now it's time to put that pipeline to real use.

In this step, you'll create a dedicated project folder and prompt Gemini CLI to design a full 5-slide Instagram carousel directly on your Paper canvas. Each slide will be its own artboard at Instagram's standard 1080x1350px portrait dimensions.

In this step, get ready to:
  • Create a project folder for your carousel content.
  • Generate a 5-slide Instagram carousel via Gemini CLI.
  • Review and refine your slides using natural language.
Set up your project folder

Before generating your carousel, you need a dedicated folder on your Desktop. This keeps your exported images and future instruction files organized in one place.

  • Exit your current Gemini CLI session by typing this command:
/quit
  • Create your project folder and move into it by running these commands:
mkdir ~/Desktop/instagram-carousels && cd ~/Desktop/instagram-carousels

Why a dedicated folder?

Gemini CLI reads a special instruction file called GEMINI.md from whatever folder you launch it in. By creating a project folder now, you'll be able to save reusable carousel preferences there in a later step.

  • Start a new Gemini CLI session inside your project folder by running:
gemini

You should see the Gemini interactive prompt appear, ready for input.

Don't see the Gemini prompt?

Make sure you ran gemini from inside the ~/Desktop/instagram-carousels folder. If you get a "command not found" error, confirm Gemini CLI is still installed globally by running npm list -g @google/gemini-cli in a separate terminal window.

Help me troubleshoot why Gemini CLI isn't starting in my project folder.

Generate the carousel with a prompt

Now for the exciting part. You'll give Gemini a detailed prompt describing exactly what carousel you want. The more specific you are about dimensions, slide structure, and style, the better the output.

  • Paste this carousel generation prompt into your Gemini CLI session:
Create a 5-slide Instagram carousel in Paper Design about 5 productivity tips for remote workers.

Requirements:
- Each slide should be a separate artboard sized 1080x1350px (Instagram portrait format)
- Slide 1: Bold hook title that stops the scroll (e.g. "5 Tips to 2x Your Remote Productivity")
- Slides 2-4: One tip per slide with a large number, short heading, and 1-2 sentence description
- Slide 5: Call to action slide ("Follow for more" or similar)
- Style: Clean, modern aesthetic with a warm color palette, bold sans-serif headings, and a subtle accent color for visual consistency
- Add a thin border or background shape to tie the slides together visually

What does this prompt do?

This prompt tells Gemini the exact slide dimensions Instagram expects (1080x1350px), the content structure (hook, tips, CTA), and the visual style. Gemini will use the Paper MCP tools (create_artboard, write_html) to build each slide directly on your canvas.

Feel free to swap the topic with anything you want to post about. The structure and style guidelines are what matter most.

Customize your topic

Replace "5 productivity tips for remote workers" with any topic you care about. Ideas: "3 beginner mistakes in Python", "4 ways to style a blazer", "5 underrated travel destinations".

  • Gemini may ask clarifying questions about colors or font preferences. Answer with whatever feels right, or just say "go with your best judgment."
  • When Gemini asks permission to call the Paper MCP tools, type yes or press Enter to approve each tool call.

Why does Gemini ask for permission?

Gemini CLI requires explicit approval before executing MCP tool calls. This is a safety feature that lets you review what the AI intends to do before it modifies your design canvas. You'll see several approval requests as it creates each artboard and writes HTML content to the slides.

This process takes 30-60 seconds as Gemini makes multiple tool calls to build all five slides.

Review and tweak your slides
  • Switch to Paper Design and look at your canvas.
  • You should see 5 separate artboards arranged on the canvas, each sized at 1080x1350px.

Each artboard represents one slide of your carousel: the hook, three content slides, and a closing call to action.

What should I see?

Your Paper canvas should now have 5 rectangular artboards. Each one contains styled text content matching your topic. The slides should have a consistent color palette and typography tying them together as a set.

If a slide looks empty or has overlapping text, that's normal for a first pass. You'll fix it in the next part.

  • If any slide needs changes, switch back to your Gemini CLI session and describe what to fix in plain English.

For example, you could type:

On slide 3, make the heading text larger and change the accent color to a softer blue.
  • Approve the MCP tool calls when Gemini asks, then switch back to Paper Design to see your changes.
  • Repeat this review-and-tweak cycle until you're happy with all 5 slides.

Slides not appearing in Paper?

Make sure Paper Design still has a file open. The MCP server only runs when a file is loaded. If you accidentally closed the file, create a new blank file and try the prompt again.

If Gemini shows an error about the MCP connection, return to your terminal and run /mcp to check the Paper server status. It should show as CONNECTED.

Help me figure out why my carousel slides aren't appearing on the Paper Design canvas.

You now have a full 5-slide Instagram carousel sitting on your design canvas, built entirely through conversation. Next up, you'll export these slides as PNG images so they're ready to upload to Instagram.

Export Your Carousel as PNGs

Your 5 carousel slides are designed and sitting on the Paper canvas. To actually post them on Instagram, you need image files you can upload.

In this step, you'll use the Paper export MCP tool through Gemini CLI to save each slide as a high-resolution PNG. Exporting at 2x scale produces crisp 2160x2700px images that look sharp on any phone screen.

In this step, get ready to:
  • Export all 5 carousel slides as PNG files at 2x resolution.
  • Verify the exported files are at Instagram-ready dimensions.
Ask Gemini to export your slides as 2x PNGs

The Paper export MCP tool can save any artboard as an image file. It supports format options (PNG, JPG, SVG) and scale overrides like 2x for high-resolution output.

  • Switch back to your Gemini CLI session (it should still be running in your terminal from the previous step).
  • Type the following prompt into Gemini CLI:
Export all 5 carousel artboards as PNG files at 2x scale.
  • When Gemini asks for permission to use the Paper MCP tools, approve the tool calls.
  • Wait for Gemini to confirm it has exported each slide.

What happens when you run this?

Gemini calls the Paper export tool once per artboard. Each call specifies PNG format and 2x scale, which doubles the pixel dimensions from 1080x1350 to 2160x2700.

This higher resolution means your carousel images stay crisp when Instagram compresses them during upload.

Export not working?

Make sure Paper Design Desktop is still open with your carousel file loaded. The MCP server only runs when a file is open.

If Gemini says it cannot find the artboards, ask it to run get_basic_info first. This refreshes its view of the canvas.

help me troubleshoot why Gemini CLI can't export my Paper Design artboards via MCP

Verify the exported PNG files

Once Gemini confirms the export, you need to check that all 5 files were created at the correct dimensions.

  • Switch to Paper Design and check the export dialog or notification area for confirmation that 5 PNGs were saved.
  • Alternatively, ask Gemini CLI to confirm the export details by typing:
Confirm the exported files. How many PNGs were exported and at what dimensions?
  • Verify that Gemini reports 5 PNG files at 2160x2700px (which is 1080x1350 at 2x scale).

Why 2x scale?

Instagram displays carousel images on high-density phone screens. A 1x export at 1080x1350 would look slightly soft on modern devices.

Exporting at 2x gives you 2160x2700px files that Instagram can compress without losing sharpness. This is the standard practice for social media image exports.

Your carousel images are ready to upload to Instagram. Next up, you'll create a reusable instruction file so you can generate new carousels with a single prompt in the future.

Save a Reusable GEMINI.md Instruction File

Your carousel images are exported and ready to post. Right now, generating a new carousel means typing all your dimension, structure, and style preferences into Gemini CLI from scratch every time.

Gemini CLI can read a special instruction file called GEMINI.md from your project folder. Once you save your preferences there, every future session automatically picks them up without you repeating anything.

In this step, get ready to:
  • Create a GEMINI.md instruction file in your project folder.
  • Add your carousel generation preferences (dimensions, slide structure, style, export workflow).
  • Test that a new Gemini CLI session reads the file automatically.
Create the GEMINI.md instruction file

Gemini CLI searches for a file named GEMINI.md in your current working directory every time you start a session. It loads the contents as persistent context that gets sent to the model with every prompt.

  • Exit your current Gemini CLI session by typing this command:
/quit
  • Create the GEMINI.md file in your instagram-carousels folder by running:
touch GEMINI.md
  • Confirm the file exists by running:
ls

You should see GEMINI.md listed alongside your exported PNG files.

Not in the right folder?

If you see files you don't recognize, navigate back to your project folder first:

cd ~/Desktop/instagram-carousels

Then run touch GEMINI.md again.

Help me find my instagram-carousels folder

Add your carousel generation preferences

This file tells Gemini exactly how you want your carousels built. It covers the MCP workflow, slide dimensions, content structure, style, and export settings.

  • Open GEMINI.md in a text editor by running:
nano GEMINI.md
  • Paste the following content into the file:
# Instagram Carousel Generator

## Instructions

When I ask you to create a carousel, follow these steps:

1. Use the Paper MCP tools to create artboards in the currently open Paper Design file.
2. Each slide should be a separate artboard sized 1080x1350px (Instagram 4:5 portrait format).
3. Use the `create_artboard` tool to make each slide, then `write_html` to add styled content.

## Default Slide Structure

- Slide 1: Bold hook title that stops the scroll
- Slides 2-4: One key point per slide with a number and short description
- Slide 5: Call to action (Follow for more)

## Default Style

- Clean, modern aesthetic with a warm color palette
- Large, bold sans-serif headings
- Subtle accent color for visual consistency
- Thin border or background shape to tie slides together

## Export

When I ask to export, use the Paper `export` tool to export each artboard as PNG at 2x scale.
  • Save the file by pressing Ctrl+O, then Enter to confirm.
  • Exit nano by pressing Ctrl+X.

What does each section do?

  • Instructions tells Gemini to use the Paper MCP tools and sets the slide dimensions to 1080x1350px.
  • Default Slide Structure defines the 5-slide format: a hook, three content slides, and a call to action.
  • Default Style sets the visual direction so slides look polished without you describing colors and fonts every time.
  • Export tells Gemini to use the Paper export tool at 2x scale whenever you ask for PNGs.

Use any text editor you prefer

If you prefer a different editor like vim or a GUI text editor, feel free to use that instead of nano. The content is what matters.

✔️ Awesome, I've got everything!

Great. Double check you saved the file before continuing.

ⓧ I'd like to double check the full code

Here is the complete GEMINI.md file for comparison:

# Instagram Carousel Generator

## Instructions

When I ask you to create a carousel, follow these steps:

1. Use the Paper MCP tools to create artboards in the currently open Paper Design file.
2. Each slide should be a separate artboard sized 1080x1350px (Instagram 4:5 portrait format).
3. Use the `create_artboard` tool to make each slide, then `write_html` to add styled content.

## Default Slide Structure

- Slide 1: Bold hook title that stops the scroll
- Slides 2-4: One key point per slide with a number and short description
- Slide 5: Call to action (Follow for more)

## Default Style

- Clean, modern aesthetic with a warm color palette
- Large, bold sans-serif headings
- Subtle accent color for visual consistency
- Thin border or background shape to tie slides together

## Export

When I ask to export, use the Paper `export` tool to export each artboard as PNG at 2x scale.
Test the instruction file with a new session

The real test is whether Gemini reads your preferences automatically. A new session started in this folder should pick up GEMINI.md without you mentioning it.

  • Start a new Gemini CLI session in your project folder by running:
gemini

Check the CLI footer. It should show that 1 context file is loaded.

  • Type this short prompt to test your saved preferences:
Create a carousel about 3 tips for better sleep.
  • Approve the MCP tool calls when Gemini asks for permission.
  • Switch to Paper Design and confirm new artboards appeared at 1080x1350px with a hook slide, 3 content slides, and a CTA slide.

You did not specify dimensions, slide structure, or style in your prompt. Gemini used your GEMINI.md preferences automatically.

How does Gemini find your instruction file?

Gemini CLI searches for GEMINI.md files in a hierarchy: first your global config at ~/.gemini/GEMINI.md, then your workspace directories. The file in your current working directory is loaded every time you start a session there.

Gemini not following your preferences?

Make sure you started the gemini session from inside ~/Desktop/instagram-carousels. If you launched it from a different directory, it won't find your GEMINI.md file.

Confirm the file is named exactly GEMINI.md (all caps, no extra extension). Run ls to double check.

Help me debug why my GEMINI.md context isn't loading

Your reusable instruction file is locked in. From now on, generating a new carousel is a one-liner. Next up, you'll take this system further with a secret mission.

Secret mission

Recreate a Carousel from a Reference Image

You've built carousels from scratch. But what if you could feed Gemini a screenshot of a carousel you admire and have it recreate the style with your own content? That's the power of multimodal AI input combined with MCP.

Clean Up Your Resources

Clean Up Your Resources

Decide whether to keep your resources running, pause them to come back later, or delete them entirely. This project runs entirely on free tools with no ongoing costs.

Resources you used:

  • Gemini CLI (free tier, 1,000 requests/day with your Google account).
  • Paper Design Desktop (free tier, 100 MCP tool calls/week).
  • Project folder at ~/Desktop/instagram-carousels.

Keep everything running

No action needed. Choose this if you're still actively building carousels or plan to generate more soon.

  • Paper Design and Gemini CLI stay installed on your machine.
  • Your GEMINI.md instruction file, exported PNGs, and reference screenshot all stay on disk.
  • To generate another carousel anytime, open Paper Design, then run gemini inside your ~/Desktop/instagram-carousels folder.

Pause - I'll come back to this later

Shut down running processes to free up system memory. Your files and configuration stay on disk for later.

  • Quit Paper Design Desktop.
  • Exit your Gemini CLI session by typing /quit if it's still running.
  • Your project folder, GEMINI.md, and all exported PNGs stay on disk.
  • To resume later, reopen Paper Design, then run gemini in your project folder.

Delete - I don't want to use this again

Remove all project resources and start fresh. This removes your exported carousels, instruction file, and the tools you installed.

  • Remove the project folder by running:
rm -rf ~/Desktop/instagram-carousels
  • Remove the Gemini CLI MCP configuration by running:
rm ~/.gemini/settings.json
  • Uninstall Gemini CLI by running:
npm uninstall -g @google/gemini-cli
  • Delete the Paper Design AppImage file from wherever you saved it (e.g. your Downloads or Applications folder).
  • If you registered the AppImage with a desktop integration tool (Gear Lever, AppImageLauncher, or AppMan), remove it from there as well.

Nice Work!

Nice Work!

Nice work! You just built a free AI-powered Instagram carousel system that turns a single text prompt into polished, export-ready slides.

Here's what you learned:

  • Connected Gemini CLI to Paper Design via MCP, giving a free AI agent direct access to a design canvas.
  • Generated and exported a full 5-slide Instagram carousel at 1080x1350px dimensions using a single prompt.
  • Created a reusable GEMINI.md instruction file that makes future carousel generation a one-liner.
  • Recreated a carousel from a reference screenshot, matching an existing style with your own content.

Ready to quiz yourself?