Claude Code + Ollama = Free Local AI Agent!

Added: 2026-05-11

[00:00:00]Claude Code plus Ollama equals a free local AI coding agent. What if I told you that you could run one of the most powerful coding tools on the planet completely offline on your own computer?

[00:00:09]No internet needed, no usage limits, no data leaving your machine, and almost nobody is talking about how easy this just got. Hey, I'm the digital avatar of Julian Goldie and I help people learn and actually use AI tools in their work without wasting time on stuff that doesn't deliver. The next few minutes I'm going to show you exactly how to pair Claude Code with Ollama. You'll learn what each tool does, how to install both, how to connect them, and the exact commands you can copy and paste right now to get this working.

[00:00:33]Stick around because near the end I'm going to share a feature that lets Claude Code run tasks on a schedule while you sleep. Let me start with what Claude Code actually is because a lot of people still think it's just a chat window. Claude Code is Anthropic's agentic coding tool.

[00:00:46]terminal. That means instead of copying code back and forth into a chat box, Claude Code can read your entire project, edit files, run commands, run your tests, and even open pull requests on GitHub for you. Works on macOS, Linux, and Windows. You can use it in your terminal, in VS Code, in JetBrains IDEs, in the desktop app, on the web, and even inside Slack. Big idea is that Claude Code understands your code base the way a teammate would. You describe what you want and it figures out which files to touch, what to change, and how to test it. Companies like Shopify, Figma, Stripe, and Notion are using it every single day to ship faster. Now here's where things get interesting. By default, Claude Code talks to Anthropic's models in the cloud, models like Sonnet, Opus, and Haiku. They're incredible. But what if you're working on a private code base you can't send to the cloud? What if you just want to learn and experiment without worrying about usage limits? That's where Ollama comes in. Ollama is a free, open-source tool that lets you run AI models directly on your own computer, models like GPT-4, Code Llama, and many others.

[00:01:41]You download them once and after that they run completely on your hardware. So you might already see where this is going. Claude Code is the brain that knows how to read code, plan changes, and edit files. Ollama lets you swap out the cloud model for a local one. And in January 2026, Ollama released version 0.14.0, which made it compatible with Anthropic's messages API. That's the API Claude Code uses to talk to its models.

[00:02:03]Plain words, this means Claude Code can now point at Ollama instead of the cloud. You get the full Claude Code experience, but the actual thinking happens on your own laptop. Let me walk you through how to set this up.

[00:02:14]Install Claude Code. If you're on macOS, Linux, or WSL on Windows, open your terminal and paste this command. curl -f ssl https://claude.ai/install.sh pipe bash. Hit enter and it installs. If you're on Windows using PowerShell, use this command instead. irm https://claude.ai/install.ps1 pipe iex. Same idea, different installer. Next, install Ollama. Go to ollama.com and download the installer for your operating system. Run it.

[00:02:46]Once Ollama is installed, you need to pull down a model. Open your terminal and type ollama pull qwen3-coder.

[00:02:52]That tells Ollama to download the qwen3-coder model to your machine.

[00:02:56]Depending on your internet, this takes a few minutes. Now, here's the magic part, connecting them. You set two environment variables in your terminal. The first one is anthropic_auth_token, and you set it equal to Ollama. Second one is anthropic_base_url, and you set it equal to http://localhost:11434.

[00:03:15]That second one is the address Ollama uses by default. Then, run Claude Code with this command. claude --model qwen3-coder. That's it. Claude Code is now running, but instead of talking to the cloud, it's talking to the model on your own machine. Now, I want to give you one critical tip here, because this is where most people trip up. Claude Code needs a large context window to work properly. Ollama recommends at least 64,000 tokens. The default in Ollama is much smaller than that. So, you want to adjust your context length settings before you start. Otherwise, Claude Code might cut off mid-task or forget what it was doing. The Ollama documentation has a full page on context length and walks you through exactly how to to it. Quick story. When I first started using AI coding tools, I was completely overwhelmed. There were so many models, so many setups, so many opinions, and most of them were wrong.

[00:03:58]That's when I created this community called AI Profit Boardroom. 2,000 members all focused on learning AI together and sharing what actually works. Taught me which workflows save time versus which ones waste it. If you're serious about using AI to improve your skills in your work, check it out.

[00:04:13]Link in the description. Okay, back to the setup. Let me talk about which models you should actually use. For local models, Ollama recommends two right now. The first is GPT-OSS-20B.

[00:04:23]The second is Qwen 3-Coder. 20B in GPT-OSS means it has 20 billion parameters, which is a fancy way of saying it's a medium-size model that runs on most modern computers without needing a beefy graphics card. If you have access to Ollama's cloud, there are options there, too. The blog post mentions GLM-4.7:cloud and Minimax-M-2.1:cloud.

[00:04:44]Cloud models always run at their full context length, so they're a good middle ground if you want better performance without giving up the workflow. You connect to a cloud model the same way.

[00:04:53]Just use Claude {dash} {dash} model GLM-4.7:cloud, for example. Now, let me show you why this combination is so useful with some real use cases. Use case one, private code bases. If you work somewhere that doesn't allow code to leave the local network, this is huge. The model reads your files, makes changes, and saves them all on your machine. Use case two, working offline. On a plane, on a train, in a cafe with bad Wi-Fi. Claude Code can still read your project, explain code, and help you fix bugs. Use case three, experimentation. Want to test different models? For Ollama, you just pull a new one and switch the {dash} {dash} model flag. Try Qwen 3-Coder for one task, GPT-OSS for another, see which fits your style. Use case four, building agents. Claude Code supports tool calling, which means the model can interact with your file system, run commands, and call APIs. You can build automated workflows that do the same task every time you run it. And speaking of automation, here's a feature most people don't know about yet. Claude Code has a {slash} loop command. You type {slash} loop, then an interval, then a prompt, and Claude Code will run that prompt on a recurring schedule. Like {slash} loop 30 minutes, and then a prompt to check your open pull requests and summarize their status. Claude Code will do that every 30 minutes in the background. You can do the same for checking GitHub issues, running research, or setting reminders. That kind of background automation is a serious time saver, especially when you're juggling multiple projects. A few tips that will save you a lot of frustration. Tip one, check your hardware. Local models use your computer's memory and processor. 20B model needs a decent amount of RAM. So, if your laptop is on the lighter side, start with something smaller. Tip two, start with simple tasks. Don't try to refactor your entire code base on day one. Ask Claude Code to explain a function or write a unit test first. Get a feel for how the local model behaves, then work your way up. Tip three, use cloud models for heavy lifting. Cloud models are still going to be faster and more capable than what runs on your laptop. It's local for privacy and learning. It's cloud for big jobs. Tip four, the Ollama docs. They have a whole page on the Claude Code integration with extra commands, headless mode for automation, and even a Telegram plugin, so you can chat with Claude Code from your phone. Where AI is right now is wild. A few years ago, the idea of running a smart coding agent fully on your own machine sounded like science fiction. Now, it's a few terminal commands away. The gap between cloud AI and local AI is getting smaller every month.

[00:07:05]It's catching up. Tools like Ollama and Claude Code are making it stupid simple to switch between them. The skill that's going to matter most is knowing how to set these tools up, which model fits which job, and how to actually use them in your daily work. That's what separates people who talk about AI from people who actually get things done with it. If you're looking to dive deeper into AI tools and actually implement them in your work, I recommend AI Profit Boardroom. 2,000 people learning how to use AI effectively. Shares real experiences. What's working? What's not?

[00:07:33]Which tools are worth your time? Which ones to skip? No hype, just solid information and practical guidance from people doing the work. Link in the description if you want to check it out.

[00:07:42]If you want the full process, SOPs, and 100 plus AI use cases like this one, join the AI Success Lab. Links in the comments and description. You'll get all the video notes from there, plus access to our community of 58,000 members who are crushing it with AI. That's it for this video. Try the setup. Start with one model. See what you can build, and I'll see you in the next one.