A free, open-source AI coding agent called Goose is challenging Anthropic's $200/month Claude Code—and it runs entirely on your laptop, no subscription required.
The $200 Problem Developers Are Refusing to Pay
Claude Code has dominated developer Twitter since New Year's Day. Engineers post videos of it building entire apps from scratch. Google's Jaana Dogan said it recreated a year's worth of work in an hour.
But there's a catch: Anthropic's pricing. The free tier offers zero access. Pro ($20/month) caps you at 10-40 prompts every five hours—exhausted in minutes of real work. The $200 Max plan gives more headroom, but even power users report hitting limits mid-session.
Enter Goose: Block's open-source answer to Claude Code. Same agentic capabilities. Same ability to write, debug, and deploy code autonomously. Zero cost. Zero cloud dependency. Zero rate limits.
What Makes Goose Different (And Why It Matters)
Goose runs on your machine. You download an open-source language model—Qwen, Llama, DeepSeek—and Goose uses it to execute coding tasks locally. Your code never leaves your laptop. You can work offline. On a plane. With no API bills.
The project has 26,100 GitHub stars and 362 contributors. The latest version shipped January 19, 2026. It's not a toy—it's a production-grade tool built by the team behind Square's payment infrastructure.
One developer who switched reported this: "I use Ollama all the time on planes—it's a lot of fun!" Another called it "exactly the tool I wish I had in 2012."
How to Set Up Goose in Under 10 Minutes
You need three things: Goose itself, Ollama (a tool for running models locally), and a language model. Here's the process:
Step 1: Install Ollama from ollama.com. One command downloads and runs a model: ollama run qwen2.5
Step 2: Download Goose from its GitHub releases page. Pre-built binaries exist for macOS, Windows, and Linux.
Step 3: In Goose Desktop, go to Settings → Configure Provider → Ollama. Set the API host to http://localhost:11434 and click Submit.
That's it. You now have an AI coding agent running on your hardware. No subscription. No usage caps. No external dependencies.
What This Means for Learners
The shift from cloud-based to local AI agents isn't just about cost—it's about control. You own the model. You control the data. You decide when to upgrade.
For anyone learning to build with AI, Goose offers a sandbox with no financial risk. Experiment with AI Agents: Build Multi-Agent Workflows without worrying about API bills. Test ideas from Claude Code Workflows: Engineering-Grade AI Skills using free, local models.
The trade-off? You need decent hardware—32GB of RAM is ideal, though smaller models run on 16GB. And local models still trail Claude Opus 4.5 on complex tasks. But the gap is closing fast. Qwen 2.5, Llama 3, and DeepSeek now benchmark near Sonnet 4 levels.
The $200-per-month era for AI coding tools may be ending. Not because the commercial products aren't good—they are. But because the open-source alternatives are now good enough for most work, and they come with something money can't buy: freedom.
