Build Smarter with Claude Code & Ollama on UNO Q

Arduino® UNO™ Q 4GB

terminal

<html><body> Turn your Arduino UNO Q into a fully functional AI development workstation. This project combines Claude Code (Anthropic's AI-powered coding assistant) with Ollama (local LLM runtime) to give you intelligent code suggestions, debugging help, and natural language programming — all running 100% offline on ARM hardware. No cloud dependencies. No API costs. No internet required. What can you do? Write code faster — Ask to generate functions, refactor logic, or explain algorithms Debug with AI — Paste error messages and get instant explanations and fixes Learn on the go — Query language syntax, library usage, or best practices without switching contexts Stay private — Your code never leaves the device. Perfect for sensitive projects or air-gapped environments Work anywhere — Train stations, planes, remote sites — if your UNO Q has power, you have AI assistance Why the UNO Q? The Qualcomm QRB2210 inside the Arduino UNO Q packs enough ARM64 compute to run lightweight LLMs at usable speeds. Tech Stack <figure><img src="https://projects.arduinocontent.cc/90d86497-b062-4a00-a20e-676f355126e0.png"/><figcaption></figcaption></figure> Hardware : Arduino UNO Q (QRB2210, ARM64, Debian) AI Runtime: Ollama v0.21.0 (CPU inference) IDE Integration: Claude Code CLI (native installer) Storage : Custom install paths to /home/arduino/ for maximum space efficiency Perfect for: Students learning to code in environments with limited internet Embedded developers working on edge devices Researchers prototyping AI-assisted workflows on ARM hardware Makers who want a personal AI assistant that fits in their backpack Ready to build? Follow the installation script and start coding smarter — offline, private, and entirely under your control. <h3>The Architecture</h3><figure><img src="https://projects.arduinocontent.cc/32bef994-b6f1-4b0b-a872-389c33107acb.png"/><figcaption></figcaption></figure> Access the board connected via USB using adb shell command <div id="gallery"><img id="gallery" src="https://projects.arduinocontent.cc/05947d87-bd52-437c-9d66-56f84189f58c.png"/><figcaption></figcaption></div> with a fresh installation, the UNO Q filesystem appears like this <figure><img src="https://projects.arduinocontent.cc/bffe92f7-1abe-4184-9169-365717b18447.png"/><figcaption></figcaption></figure> App Lab uses the the /dev/mmcblk068 for downloading and create containers and has around 2Gb of free space, instead the /dev/mmcblk069 has plenty of space , around 17Gb, so we will use this partition to install ollama code and the models downloaded with it <h3>Install Claude on UNO Q</h3>first we export the path that Claude will use for installing using /home/arduino, that is the mount point available. <pre>export CLAUDE_HOME=/home/arduino </pre><pre>export PATH="$CLAUDE_HOME/bin:$PATH" </pre> then we will add a permanent declaration tha will export the path at every reboot <pre>echo 'export CLAUDE_HOME=/home/arduino' &gt;&gt; ~/.bashrc </pre><pre>echo 'export PATH="$CLAUDE_HOME/bin:$PATH"' &gt;&gt; ~/.bashrc </pre> Now Claude knows where to install so we can launch the standard command used to download the installer. <code class="code">curl -fsSL https://claude.ai/install.sh | bash</code> <figure><img src="https://projects.arduinocontent.cc/25253bae-0272-4214-891e-140ad9f604c7.png"/><figcaption></figcaption></figure> so now with the command <code class="code">which -a claude</code>we will se that is installed correctly <figure><img src="https://projects.arduinocontent.cc/37a19b5e-9d00-46bb-bd56-a9bf8157122b.png"/><figcaption></figcaption></figure> <h3>Test running Claude</h3>let's start Claude to check that everything works as expected <figure><img src="https://projects.arduinocontent.cc/5b1ee4a4-0830-4a59-acb9-25df874331f5.png"/><figcaption></figcaption></figure> <h3>INSTALL OLLAMA</h3>Now it's time to install OLLAMA. we will install it in the same way we did for Claude: Download the binary for ARM64 <pre>curl -L https://github.com/ollama/ollama/releases/download/v0.21.0/ollama-linux-arm64.tar.zst -o /home/arduino/ollama.tar.zst </pre><pre>cd /home/arduino </pre><pre>tar --use-compress-program=zstd -xf ollama.tar.zst </pre><pre>chmod +x /home/arduino/ollama </pre>Let's create the Directory that will contain the models: <pre>mkdir -p /home/arduino/.ollama/models </pre> let's update the PATH <pre>echo 'export OLLAMA_MODELS=/home/arduino/.ollama/models' &gt;&gt; /home/arduino/.bashrc </pre><pre>echo 'export PATH="/home/arduino:$PATH"' &gt;&gt; /home/arduino/.bashrc </pre><pre>source /home/arduino/.bashrc </pre> Now we need to create the systemd service that will be used to launch OLLAMA at restart sudo nano /etc/systemd/system/ollama.service Copy and paste the following text <pre>[Unit] </pre><pre>Description=Ollama Service </pre><pre>After=network-online.target </pre><pre>[Service] </pre><pre>ExecStart=/home/arduino/ollama serve </pre><pre>Environment="OLLAMA_MODELS=/home/arduino/.ollama/models" </pre><pre>User=arduino </pre><pre>Group=arduino </pre><pre>Restart=always </pre><pre>RestartSec=3 </pre><pre>[Install] </pre><pre>WantedBy=default.target </pre><pre>	 </pre> <figure><img src="https://projects.arduinocontent.cc/5cc42a76-391b-4062-b36b-e7a570e2b73b.png"/><figcaption></figcaption></figure> now press <code class="code">CTRL O </code>to write the file (write return to confirm writing) and <code class="code">CTRL X </code>to exit the editor <figure><img src="https://projects.arduinocontent.cc/fed961ab-8789-475c-a470-f9330b39319c.png"/><figcaption></figcaption></figure> now we can check if everything is correctly setup <pre>sudo systemctl daemon-reload </pre><pre>sudo systemctl enable ollama </pre><pre>sudo systemctl start ollama </pre> ok, if you don't read nothing everything is good to go! now test with <pre>ollama --version </pre> <figure><img src="https://projects.arduinocontent.cc/eeae226e-4c62-468e-a24e-48293dd3c50d.png"/><figcaption></figcaption></figure> if is ok let's remove the installation file to free up some space <pre>rm /home/arduino/ollama.tar.zst </pre> Let's now install the model with OLLAMA, this is a quick comparison <figure><img src="https://projects.arduinocontent.cc/740d42a2-caa2-4bee-963e-ba4a003a72ab.png"/><figcaption></figcaption></figure> so let's start with the llama3.2:3b launch this command <pre>ollama </pre><pre>ollama run </pre><pre>	llama3.2:3b </pre><pre>	 </pre>after downloading the model (it's around 2 Gb) you'll be able to interact with the model from within Ollama <figure><img src="https://projects.arduinocontent.cc/74ca94c6-85f4-4fd3-9937-34722f3fee1c.png"/><figcaption></figcaption></figure> let's start to interact with Ollama: <figure><img src="https://projects.arduinocontent.cc/4e388785-9f11-4f18-b569-951f92969895.png"/><figcaption></figcaption></figure>it works!! Last part: connect ollama with claude code: launch <pre>ollama launch claude </pre>go down with the arrow and select llama3.2:3b as model <figure><img src="https://projects.arduinocontent.cc/c6e66a96-2a43-43eb-bc30-13c84ac557a3.png"/><figcaption></figcaption></figure> if you are getting issues, reforce exporting path <pre>rm /home/arduino/.local/bin/claude </pre><pre>ln -s /home/arduino/.local/share/claude/versions/2.1.117 /home/arduino/.local/bin/claude </pre><pre>chmod +x /home/arduino/.local/bin/claude </pre><pre>echo 'export PATH="/home/arduino/.local/bin:$PATH"' &gt;&gt; ~/.bashrc </pre><pre>source ~/.bashrc </pre> and relaunch <pre>ollama launch claude </pre> now, you should see the claude code CLI interface initializing. select the theme (dark, light etc) then trust the folder and .... you have now Claude code running on your UNO Q using a free model! <figure><img src="https://projects.arduinocontent.cc/4758e274-d648-4062-b890-1effa4165eb8.png"/><figcaption></figcaption></figure> <figure><img src="https://projects.arduinocontent.cc/e6b40810-a117-494b-a6ae-c010387ae7ec.png"/><figcaption></figcaption></figure> <figure><img src="https://projects.arduinocontent.cc/f75cf3e5-8b07-4b4c-b9e6-120c04c47f56.png"/><figcaption></figcaption></figure> within the first prompts, the model can take longer time to answer, if you want, try with different models to find the most suitable model for your needs. </body></html>

Build Smarter with Claude Code & Ollama on UNO Q

Your personal AI coding assistant running 100% locally