Imagine ditching expensive AI coding tools and embracing a completely free, local alternative. Sounds too good to be true, right? Well, that's exactly what I set out to explore by testing a combination of Goose, an open-source agent framework, and Qwen3-coder, a powerful coding-focused language model. But here's where it gets controversial: Can these free tools truly replace pricey options like Claude Code? Let's dive in and find out.
Jack Dorsey, the tech visionary behind Twitter (now X), Square (now Block), and Bluesky, sparked curiosity in July with a cryptic tweet: "goose + qwen3-coder = wow." This simple equation ignited interest in both tools, particularly as a potential free alternative to Claude Code. Goose, developed by Block, is an open-source agent framework, while Qwen3-coder, from Alibaba, is a coding-centric large language model. Together, they promise a fully local, cost-free coding solution. But does this combination live up to the hype?
And this is the part most people miss: Setting up this local AI stack isn't as daunting as it seems, but it does require a robust machine. Here’s how I got it running on my Mac, though the process is similar for Windows or Linux users.
Getting Started: Downloading the Software
First, you’ll need to download Goose from GitHub and Ollama, an LLM server, from its official site. I initially installed Goose first, only to realize I needed Ollama to make it work—a rookie mistake! So, pro tip: Install Ollama first. Once Ollama is up and running, you’ll download the Qwen3-coder model directly from within it. Keep in mind, this model is a hefty 17GB, so ensure your machine has ample storage.
Installing Ollama and Qwen3-coder
After downloading Ollama, double-click the installer and launch the app. You’ll see a chat-like interface with a default model (likely gpt-oss-20b). Click on it to select Qwen3-coder:30b, a model optimized for coding with 30 billion parameters. The model won’t download until you prompt it—I simply typed "test" to kickstart the process.
One of the standout benefits here is that everything runs locally. No data is sent to the cloud, giving you full control over your work. To ensure Ollama is accessible to other applications, head to Settings and enable "Expose Ollama to the network." I also set my context length to 32K, though you can adjust this based on your machine’s capabilities.
Installing Goose
With Ollama ready, it’s time to install Goose. Choose the version that suits your system—I opted for the MacOS Apple Silicon Desktop version. Upon launching Goose for the first time, you’ll be greeted with a Welcome screen. Navigate to "Other Providers" and configure Ollama as your LLM provider. Here’s where I initially stumbled: "Configure Ollama" doesn’t mean tweaking Ollama itself but rather setting up the connection between Goose and Ollama. Select the Qwen3-coder:30b model, and you’re good to go.
Putting Goose to the Test
I tested Goose with a standard challenge: building a simple WordPress plugin. On the first attempt, it generated a plugin, but it didn’t work. After several retries and corrections, Goose finally succeeded on the fifth try. While this might seem like a lot of attempts, there’s a silver lining: agentic coding tools like Goose and Claude Code work directly on the source code, meaning each correction improves the codebase.
But here’s the real question: Can this free setup truly compete with paid alternatives? My initial impressions are promising. Running on an M4 Max Mac Studio with 128GB of RAM, I didn’t notice a significant difference in performance compared to cloud-based tools like Claude Code or OpenAI Codex. However, this was just a simple test. The true test will come when I tackle a larger project.
Final Thoughts and Your Input
While Goose and Qwen3-coder show potential, they’re not without quirks. Setup is straightforward, but accuracy and retries remain areas for improvement. Here’s where I need your input: Have you tried running a coding-focused LLM locally? How did it go? Did you notice a difference in performance compared to cloud-based tools? Share your thoughts in the comments below.
Stay tuned for my next article, where I’ll delve deeper into the roles of Goose, Ollama, and Qwen3-coder in the AI agent coding process. And don’t forget to follow my journey on social media or subscribe to my newsletter for more updates!
