I tried a local, open-source, and completely free rival to Claude Code - how it went
ZDNET's key takeaways
- Free AI tools Goose and Qwen3-coder may replace a pricey Claude Code plan.
- Setup is straightforward but requires a powerful local machine.
- Early tests show promise, though issues remain with accuracy and retries.
Jack Dorsey, the founder of Twitter (now X), Square (now Block), and Bluesky (still Blue), posted a cryptic statement on X, saying "goose + qwen3-coder = wow".
Goose, developed by Dorsey's company Block, is an open-source agent framework, similar to Claude Code. Qwen3-coder is a coding-centric large language model similar to Sonnet-4.5. Both are free.
Together, they can potentially create a fully free competitor to Claude Code. But can they? I decided to find out.
This is the first of three articles that will discuss the integration of Goose (the agent framework), Ollama (an LLM server), and Qwen3-coder (the LLM).
In this article, I'll show you how to get everything working. In the next article, I'll provide a more in-depth understanding of the roles each tool plays in the AI agent coding process. Finally, I'll attempt to use these tools to build a fully powered iPad app.
I'm building this on my Mac, but you can install all three tools on your Windows or Linux machine.
Downloading the software
Start by downloading Goose and Ollama. Later, download the Qwen3-coder model from within Ollama:
- Download Goose from GitHub.
- Download Ollama.
I initially downloaded and installed Goose first, but couldn't get it to talk to Ollama. I hadn't yet downloaded and set up Ollama. My recommendation is to install Ollama first.
Installing Ollama and Qwen3-coder
Install Ollama, then double-click the installer. Once loaded, you'll see a chat-like interface. Choose a model, such as Qwen3-coder:30b, which is a coding-optimized model with about 30 billion parameters.
Note that the model won't download until it's forced to answer a prompt. I typed "test" and the model downloaded.
Finally, set your context length to 32K. I have 128GB of RAM, so if I start to run out of context, I'll boost it.
Installing Goose
Next, install Goose. Run the installer and choose the MacOS Apple Silicon Desktop version.
After launching Goose for the first time, you'll see a Welcome screen. Go to the Other Providers section and click Go to Provider Settings.
Configure your connection to Ollama, choosing qwen3-coder:30b as the model.
Taking Goose for a spin
Type a prompt into the prompt area. Set Goose to work from a temporary folder and specify the model.
As a test, I built a simple WordPress plugin. Goose/Qwen3 failed initially, but after repeated corrections, it eventually got it right.
First impressions
I was disappointed it took Goose five tries to get my test to work. Other free chatbots got it right on the first try.
But agentic coding tools like Claude Code and Goose work on the actual source code, so repeated corrections improve the codebase.
My colleague Tiernan Ray found performance unbearable on his 16GB M1 Mac. On my M4 Max Mac Studio with 128GB of RAM, I found overall performance quite good.
Stay tuned for further analysis on whether this free solution can replace expensive alternatives like Claude Code's Max plan or OpenAI's Pro plan.
