DIY Mini PC Runs My Personal AI Agent

The world is moving incredibly fast these days because of Large Language Models (LLMs). Over the past year we have seen how powerful these models have become — from writing code and answering questions to building entire applications.

Now things are getting even more interesting with projects like OpenClaw, which allow LLMs to actually control computers and servers to perform real tasks. These AI agents can do many of the things we normally do on our computers every day — clearing emails, managing calendars, checking flight bookings, writing code, managing servers, designing CAD models, developing games, and much more.

In many cases, these systems can perform complex workflows with very little human intervention.

For someone like me who loves technology, this feels almost like magic. ✨

But at the same time, it can also be a little scary.

Giving an AI agent full access to your personal computer or production systems can be risky. It could potentially access sensitive files, modify important settings, or expose private information such as emails, accounts, or even financial data. Since tools like OpenClaw are still in a very early stage, security policies and safeguards are still evolving. Because of this, many experts currently recommend not running AI agents directly on your primary computer.

A better approach is to run these agents on a separate secondary system — something that can safely connect to the internet and perform tasks without risking your main machine.

Some people are using systems like the Mac mini or the Raspberry Pi for this purpose. However, the Mac mini can be quite expensive if you just want to experiment with AI agents, and the Raspberry Pi can be too slow for heavier LLM workloads.

So I decided to build something that sits right in the middle between cost and performance.

The result is this $200 Portable AI Agent Mini PC, built around the powerful LattePanda IoT A — a compact single-board computer powered by the Intel N150 processor. What makes the LattePanda especially interesting is that it also includes an onboard RP2040 microcontroller, which opens the door for creative hardware integrations alongside the main computer.

To make this system more practical and fun to use, I added several extra features:

🔹 UPS battery hat to make the system portable

🔹 WiFi card and external antennas for reliable connectivity

🔹 Status LEDs for quick visual feedback

🔹 A 0.96" OLED display that shows the current AI agent state

The OLED display provides live feedback about what the LLM is doing, such as:

• Idle

• Thinking

• Loading

• Success

• Error

• and other operational states

The fun part is that these indicators are controlled directly by the AI agent itself, giving the system a physical interface that reflects what the AI is doing internally.

Finally, everything is housed inside a simple custom 3D-printed enclosure, making the whole setup compact, portable, and easy to place anywhere on your desk.

In this Instructable, I will show you how to build your own portable AI Agent Mini PC, install OpenClaw, and create a system that can run powerful AI agents while keeping your main computer safe.

Let’s get started. 🚀

Supplies

• 1×LattePanda IoT A
• 1×LattePanda IOTA Active Cooler
• 1×LattePanda UPS Expansion
• 1×Intel BE200 Wi‑Fi 7 Bluetooth 5.4 M.2 E‑Key Wireless Card
• 1× 0.96" OLED Display
• 3× 18650 Li-ion Batteries
• 2× Green LEDs
• 1× Red LED
• 1× Yellow LED

Notes

The required screws & Standoffs come included with the LattePanda IOTA and the UPS Expansion board.

The RTC battery is also included with the LattePanda IOTA.

The UPS connector cable come included with the UPS Expansion board.

Antenna and Antenna connector wire come included with the Wireless Card.

Tools

• 3D Printer
• Soldering Station
• Screwdriver Set

Step 1: About the LattePanda IoTA

LattePanda IoT A is a palm-sized, low-power embedded x86 single-board computer powered by the Intel Twin Lake N150 quad-core processor — the next-generation successor to Alder Lake N100 with long lifecycle support.

Compared with boards like the Raspberry Pi 5, the IOTA delivers 2–3× CPU performance and 10–30× GPU performance, making it capable of running demanding workloads such as edge AI, TinyML, and embedded AI applications.

The board is available in two configurations:

• 8GB RAM / 64GB eMMC
• 16GB RAM / 128GB eMMC

It supports both Windows and Linux, and the system’s adjustable TDP (6W–15W) allows developers to balance power consumption and performance depending on the application. This flexibility makes it suitable for robotics, Industrial IoT (IIoT), and edge computing systems.

Another major advantage of the IOTA platform is its expansion ecosystem, which includes modules like:

• UPS expansion boards 🔋
• PoE modules 🌐
• M.2 M-Key storage expansion 💾
• 4G LTE connectivity 📡

These options allow developers to build specialized systems such as:

• Robotic control gateways
• Machine-vision edge devices
• AI-powered robots
• Digital signage systems
• Autonomous logistics platforms
• Mobile data loggers

The board also supports ultra-thin eDP touchscreens, making it ideal for handheld instruments, portable medical devices, field research kits, and off-grid monitoring systems.

One of the most interesting features of the IOTA is the integrated Raspberry Pi RP2040 co-processor. This microcontroller can handle real-time tasks such as sensor reading, motor control, and hardware I/O, while the main Intel processor focuses on heavier workloads like AI models and automation.

This combination of x86 computing power + microcontroller-level control creates a powerful hybrid platform for robotics, embodied AI, and LLM-driven automation at the edge.

Why This Is Useful for Our Project 🤖

For this project, we need a system that can run an AI agent continuously, connect to the internet, and safely perform automation tasks without risking our main computer.

The LattePanda IoT A is perfect for this because:

🔹 Enough CPU power to run local LLM tools and AI agent frameworks smoothly

🔹 x86 architecture allows easy installation of Linux tools and AI software

🔹 Built-in RP2040 lets us control hardware like LEDs and displays directly from the AI agent

🔹 M.2 expansion support allows adding fast Wi-Fi and other modules

🔹 Low power consumption makes it ideal for portable and always-on systems

🔹 Compact size makes it easy to fit inside a custom 3D-printed enclosure

In this build, the LattePanda acts as the brain of the AI Agent Mini PC, running the automation software while the RP2040 handles the status LEDs and OLED display that show what the AI agent is doing in real time.

This makes it possible to build a dedicated, portable AI agent system that is powerful, expandable, and safe to run independently from your main workstation.

Step 2: About OpenClaw

OpenClaw is an open-source AI agent framework that lets you create and run autonomous assistants on your own hardware or server. Unlike traditional AI chatbots that only answer questions, OpenClaw’s agents can actually do things — like automate workflows, interact with applications and files, and execute multi-step tasks on your behalf.

At a high level, OpenClaw works by connecting a large language model (LLM) with local system resources, communication platforms, and “skills” that define what the agent is allowed to do. When you send it a message through a platform like WhatsApp, Telegram, Discord, Slack, Signal, or iMessage, the agent interprets your intent, plans actions, and executes them — returning results through the same channel.

Here’s how it functions in practice:

• 🌐 Local AI Execution – OpenClaw installs on your machine or server and runs persistently, rather than as a one-off browser chatbot session.
• 🛠 Tool Integration – It can access system tools, scripts, browser interactions, and APIs to perform actions.
• 📌 Persistent Memory – OpenClaw maintains conversation context across sessions, so it remembers preferences, workflows, and ongoing tasks.
• 🧩 Skills & Plugins – The system supports a large ecosystem of community-created skills that extend its capabilities — from task scheduling to file management and beyond.
• 🌍 Multi-Channel Interaction – You can communicate with your agent through messaging apps you already use, turning chats into powerful task commands.

Because OpenClaw can execute real actions (not just reply), it effectively turns your machine into a 24/7 autonomous assistant that interfaces with tools, files, and systems. This ability to act — rather than just converse — is what distinguishes it from ordinary AI chat models.

However, running an agent with deep system access also carries security considerations, especially if it has broad permissions or connects to sensitive resources — which is one reason projects like this portable Mini PC isolate the agent on dedicated hardware.

Step 3: CAD & 3D Printing

Since we have finalized all the components and functions of our system, the next step is to design an enclosure and 3D print it to hold everything together.

I designed the enclosure in Autodesk Fusion 360. To ensure accurate fitting, I imported the CAD models of the LattePanda IoT A and the LattePanda UPS Expansion from their official product files and datasheets. For the components whose models were not available, such as the OLED display and LEDs, I created simplified models based on their dimensions.

The enclosure is designed as a four-part assembly:

Housing

This is the main body that holds all the components. It includes cutouts for the ports of the LattePanda IOTA as well as an opening for the UPS charging port.

Cover

The cover sits on top of the housing and includes mounting openings for the OLED display and status LEDs. It is designed with a snap-fit mechanism, so it securely attaches to the housing without requiring any screws.

Button Extensions

Since the power and reset buttons on the IOTA board sit inside the enclosure, I designed small button extension pieces that allow these buttons to be accessed from outside the case.

GPIO Cover

The bottom of the LattePanda exposes GPIO pins, so I added a small protective cover.

Download and Print

You can either:

• Download the STL files and directly 3D print the parts
• Or download the Fusion 360 source file if you want to modify the design to fit your own requirements

Printing Settings

I printed the enclosure using PLA filament, which is easy to print and works well for prototyping.

My configuration:

• Material: PLA
• Color: White for the housing
• Cover: White + Black combination for better aesthetics
• Layer Height: 0.2 mm
• Infill: 20–25%

This results in a clean, compact enclosure that neatly houses all the components while keeping the device portable and visually appealing.

IoTA_Fusion_360

Step 4: Standoffs Assembly

In this step we will assemble the standoffs that will hold the boards securely inside the enclosure.

Take the M3 screws and M3 × 18 mm standoffs and install them on both the boards:

• Place 4× M3 × 18 mm standoffs onto the mounting holes of the LattePanda IoTA.
• Secure them using M3 screws.
• Repeat the same process for the LattePanda UPS Expansion board.
• Make sure all 4 standoffs on each board are firmly tightened so the boards remain stable when installed in the enclosure.

These standoffs create the correct spacing between the boards and the enclosure, preventing short circuits and allowing airflow for cooling.

Step 5: Wireless Card Assembly

In this step we will install the wireless connectivity module.

Take the Intel BE200 Wi‑Fi 7 Bluetooth 5.4 M.2 E‑Key Wireless Card and first connect the Wi-Fi and Bluetooth antenna cables to the small U.FL connectors on the card. These connectors are delicate, so press them down gently until they snap into place.

Next, locate the M.2 E-Key slot on the LattePanda IoT A.

• Insert the wireless card into the M.2 E-Key slot at a slight angle.
• Push the card down so it sits flat against the board.
• Use the small mounting screw that came with the wireless card to secure it in place.

Once installed, the board will support high-speed Wi-Fi and Bluetooth connectivity, allowing the AI agent system to connect to the internet and communicate with external services and devices.

Step 6: LED Connection & Assembly

In this step we will install and wire the status LEDs that visually indicate the system state.

First, take the LEDs and insert them into the dedicated slots on the enclosure cover. The holes in the cover are designed to hold the LEDs securely in place.

Once the LEDs are positioned, follow the connection diagram and connect them to the appropriate pins.

🟢 IOTA Status LEDs

These LEDs indicate the power state of the system.

Power LED (Green)

• Connect the positive leg to the S0 pin
• Connect the negative leg to GND

Sleep LED (Yellow)

• Connect the positive leg to the S3 pin
• Connect the negative leg to GND

These signals come from the LattePanda IoT A, allowing the LEDs to show whether the system is powered on or in sleep mode.

🔋 UPS Status LEDs

These LEDs show the battery charging status from the LattePanda UPS Expansion.

Charging LED (Red)

• Connect to R+ and R−

Charged LED (Green)

• Connect to G+ and G−

Once all the LEDs are connected, they will provide quick visual feedback about the system’s power and charging state without needing to connect a monitor.

Step 7: OLED Connection & Assembly

• Take the 0.96" OLED display and place it into the dedicated slot on the enclosure cover.
• Align the mounting holes of the OLED with the holes on the cover.
• Secure the display using 4 × M2 × 5 mm screws.

Make sure the screen is properly aligned with the opening so it is clearly visible from the outside.

Next, connect the OLED to the I²C interface of the LattePanda IoT A.

Follow these wiring connections:

OLED PinIOTA Pin

• GND - GND
• VCC - 3.3V
• SCL - GPIO5
• SDA - GPIO4

The OLED will communicate with the system through the I²C interface, allowing the software to display different AI agent states

Step 8: LattePanda Assembly

LattePanda Iota features a 3V Real-Time Clock(RTC) battery connector. This allows the board to maintain its internal clock even when the main power supply is completely disconnected.

Take the button extension pieces and insert them into the button openings on the cover. These extensions allow you to press the onboard buttons from outside the case.

Next, take the LattePanda IoT A board and carefully position it inside the cover.

• Align the Power and RST (Reset) buttons on the board with the button extension pieces.
• Make sure the button extensions sit correctly on top of the board buttons.
• Gently place the board into position so everything lines up properly.
• Also screw the wireless card external antenna connector on to the housing.

Step 9: UPS Connection & Assembly

In this step we will connect the UPS module to the main board.

Take the UPS connector cable and connect the LattePanda UPS Expansion to the LattePanda IoT A.

• Locate the UPS connector port on the IOTA board.
• Connect the UPS cable between the IOTA and the UPS board.
• Carefully check the connector orientation before inserting it.

⚠️ Important:

Make sure to follow the correct connector orientation shown in the reference image. Do not force the connector in the reverse direction, as this may damage the boards and potentially harm the entire system.

Once connected correctly, the UPS will provide battery backup power, allowing the system to run from the 18650 batteries and continue operating even when external power is disconnected.

Step 10: Final Assembly

Secure the IoTA and UPS board:

• Insert 8 × M3 × 8 mm screws from the bottom of the housing.
• Tighten them into the standoffs attached earlier to firmly secure both boards in place.

After the boards are secured:

• Carefully route all the wires so they do not get pinched.
• Snap the cover onto the housing, ensuring it locks properly into place.
• Place the GPIO cover onto the opening at the bottom of the housing to protect the exposed GPIO pins
• Take the external antennas and screw them onto the antenna connectors of the wireless card.

Once completed, your Portable AI Agent Mini PC enclosure is fully assembled and ready for the next steps.

Step 11: First Boot & RP2040 Setup

Now that the hardware assembly is complete, it's time to power on the system and set up the RP2040 microcontroller.

First, make sure the UPS battery is fully charged before powering on the system. Then press the Power button to boot the LattePanda IoT A.

For this project, I installed Ubuntu 24.04.4 LTSas the operating system.

Install Thonny:

Once the system boots into Ubuntu:

• Open the App Store.
• Search for Thonny.
• Install the application.

Thonny will be used to program the onboard RP2040 microcontroller.

Configure the Interpreter

After installing Thonny:

• Open Thonny.
• Go to Tools → Options → Interpreter.
• Select MicroPython (RP2040) as the interpreter.
• Leave the port set to "Detect port automatically".
• Click OK.

When you return to the Thonny window, the MicroPython interpreter should appear in the shell, indicating that the RP2040 is connected and ready.

Install the OLED Library

Next we need to install the library used to control the OLED display.

• Go to Tools → Manage Plugins.
• Search for SSD1306.
• Install the plugin.

This library allows the RP2040 to communicate with the 0.96" OLED display.

Upload the Project Files

Now download the project files from GitHub:

Repository:

https://github.com/MukeshSankhla/Portable-Ai-Agent-PC

• Download the repository.
• Extract the files.
• Open Thonny.
• Copy and paste the following files into the editor:
• main.py
• openclaw_ui.py
• Click Save As and save both files with the same names on the RP2040.

After saving the files, your setup should look similar to the reference images provided.

Open main.py and click Run.

If everything is configured correctly, the OLED display will turn on and start showing the interface content.

This confirms that the RP2040, OLED display, and hardware connections are working correctly.

Note:

The LattePanda IoT A already comes with the MicroPython bootloader installed on the RP2040.

If for any reason you encounter an error or the RP2040 is not detected, you can reinstall the bootloader by following the official guide:

https://docs.lattepanda.com/content/iota_edition/rp2040_programming/

At this point, your hardware interface system is fully operational.

Step 12: OpenClaw Setup

Now that the hardware and RP2040 interface are ready, we can install the AI agent framework.

Installing OpenClaw is very simple. Open a terminal and run the following command:

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

Then sit back and relax — the installer automatically downloads and configures everything required for OpenClaw to run.

Once the installation finishes, you will see the OpenClaw security prompt in the terminal.

Select Yes and proceed.

Setup Mode

You will be asked to choose a setup mode.

• Select: Quick Start

Selecting the AI Model

Next, OpenClaw will ask you to select a model provider. Some available options include:

• OpenAI
• Anthropic
• DeepSeek
• Google Gemini
• and more...

Most of these models require usage credits. For this project I am using DeepSeek because it has one of the lowest running costs and allows a minimum $2 top-up, which is perfect for experimenting with AI agents.

To set this up:

• Go to https://platform.deepseek.com/api_keys
• Log in to your account.
• Add balance to your account.
• Create a new API key.
• Copy the API key and paste it into the terminal when prompted.

After entering the API key, you will be asked to choose the model.

For this project I selected: deepseek-chat

Different models have different pricing and token consumption, so make sure to review them before choosing.

Selecting a Communication Channel

Next, OpenClaw will ask you to select a communication channel.

These channels allow you to interact with your AI agent remotely using apps you already use every day, such as:

• WhatsApp
• Telegram
• Slack
• Discord
• and more

For this project I chose Discord, because it does not require linking a phone number and can be created using just an email account.

For security reasons, I recommend using a secondary email account so that if anything gets compromised it does not affect your primary accounts.

Here is the process:

• Create a new email account.
• Create a new Discord account using that email.
• Create a new Discord server and channel where you will communicate with the AI agent.

You can follow this tutorial to set up the Discord bot and channel:

https://youtu.be/VfmG9Q69p1o

Once your bot is created:

• Go to your Discord Bot page.
• Click Refresh Token.
• Enter your account password.
• A new Discord Bot Token will be generated.
• Copy the token and paste it into the terminal.

For Discord Channel Access, select: Allow All Channels

Additional Setup Options

OpenClaw will then ask a few additional configuration questions.

You can skip this step for now.

Select all available options.

Choosing the Interface (Hatch)

Finally, you will be asked how you want to interact with the AI agent.

Two options are available:

Web Hatch

• Opens a web interface in your browser where you can chat with the AI agent and configure settings.

TUI (Terminal UI)

• Allows you to chat with the AI agent directly from the terminal.

For this build, I selected Web Hatch.

And that's it!

Your AI Agent Mini PC is now fully configured, and you are ready to have your first conversation with your AI agent through Discord or the web interface.

If you exit the setup terminal for some reason and want to start the setup process again, simply run:

openclaw onboard

If you want to remove OpenClaw and perform a fresh installation, you can run:

openclaw uninstall

Step 13: Interacting With the AI Agent

Now that everything is set up, it's time to start interacting with your AI agent.

For the first conversation, it is recommended to define the agent’s identity and responsibilities. This helps the AI understand how it should behave and what kind of tasks it should assist with.

You can start by sending a short message describing:

• The name of the AI agent
• Its role or duties
• A little about yourself
• The type of work you do
• Your preferred work style

For example, you can write a short paragraph introducing yourself and defining how the agent should assist you. This helps the AI provide more relevant and structured responses.

Chat Through Discord

If you selected Discord as the communication channel, you can simply chat with the agent inside your Discord server.

To get a response from the bot, you need to mention it using its handle:

@bot_name your message here

Once mentioned, the AI agent will process the request and respond in the channel.

Step 14: Making the AI Agent Perform a Task

Now that we have already configured the RP2040 and the OLED display to show the status of the AI agent, we will make the AI agent use those OLED functions by itself.

The goal here is to let the AI agent read the RP2040 code and create its own skill so it can always update the OLED display with its current state and task feedback.

Since the OLED control logic already exists in the files we uploaded earlier (main.py and openclaw_ui.py), we can simply ask the AI agent to analyze the code and build a skill around it.

Sending the Instruction

All you need to do is send the following message to your AI agent:

“Can you go to Thonny and check the main.py code, and its supported module. And build a skill around the same OLED functions where you should always show the status/feedback's of your tasks on the OLED using the same, remember this and use this skill always, also change the status from Ideal to some relevant OLED message before performing any task indicating you are busy.”

What Happens Next

Once you send the message, the OpenClaw AI agent will begin processing the request.

You will see the agent:

• Navigate through the RP2040 project directory
• Read and analyze the main.py code
• Try & understand the available OLED functions
• Create a new skill that uses those functions

After creating this skill, the AI agent will start using it automatically.

Result

From this point onward, the AI agent will try to update the OLED display whenever it performs a task, showing states such as:

• Idle
• Thinking
• Loading
• Success with message
• Error with error message

This creates a physical feedback interface for the AI agent, allowing you to see what the agent is doing directly on the device without opening the terminal or web UI.

Step 15: Conclusion

And with that, your Portable AI Agent Mini PC is complete! 🎉

With this setup, you now have a dedicated system running an AI agent that can interact with you through apps like Discord or through the web interface provided by OpenClaw.

You can experiment with many different tasks using your AI agent. For example, you can ask it to:

• Write scripts to run the LattePanda IoT A in headless mode
• Build simple applications or dashboards
• Help with design work or coding tasks
• Manage emails and documents
• Automate Excel workflows
• And much more...

Because the system is running on a separate machine, you can safely experiment with these powerful AI tools without exposing your main computer to potential risks.

If you want to learn more about OpenClaw and its capabilities, I highly recommend watching this video:

https://youtu.be/T-HZHO_PQPY

It gives a good overview of what AI agents can do and how they are evolving.

My Opinion

Personally, I’m still not completely sure what the long-term future of AI and AI agents will look like.

What I do see, however, is that the entire world is moving rapidly toward AI. Almost every major technology company is investing billions of dollars into AI development. It has become one of the highest priorities in the tech industry.

At the same time, there are still many important global challenges — such as environmental protection, food production, employment, and safety of living beings — that also deserve attention.

Over the past two years, the shift in everyday life because of AI has been remarkable. Tools powered by AI are already changing how people work, learn, and build technology.

So whether we like it or not, the best thing we can do is understand it and learn how to use it responsibly.

Soon we will likely see AI systems integrated deeply into robotics and automation, giving them access to the physical world, not just the digital one.

But it’s still worth asking ourselves some important questions:

• Can AI grow food? 🌱
• Can it heal and protect our environment? 🌍
• Can it make our societies safer and more humane?

And if it can… at what cost?

These are questions the world will have to answer as AI continues to evolve.

For now, the best thing we can do is keep building. Projects like this help us experiment, learn, and understand the possibilities of this powerful technology in a safe and controlled way.