Transcription
[Music] [Applause] Hi, welcome to another video.
So, GitHub has launched their C-pilot CLI. This is supposed to be their own CLI agent that competes with tools like Claude Code and Codeex, but it is based on the same copilot agent architecture that powers Copilot in VS Code or Copilot agent in GitHub. I haven't used Copilot for a while, but I have heard great things about how good it is. Especially these days, many companies are giving their employees C-Pilot subscriptions. So, it's good to see that they can now also code using the terminal and probably script some stuff and things like that.
They say that Copilot CLI is now in public preview. "We're bringing the power of GitHub Copilot coding agent directly to your terminal. With GitHub Copilot CLI, you can work locally and synchronously with an AI agent that understands your code and GitHub context. It allows you to work with the C-pilot coding agent directly in your command line. No context switching required. You can access your repositories, issues, and pull requests using natural language, all authenticated with your existing GitHub account. You can build, edit, debug, and refactor code with an AI collaborator that can plan and execute complex tasks. Take advantage of the fact that the coding agent ships with GitHub's MCP server by default and supports custom MCP servers to extend capabilities. It is also different from some other coding agents that are always on auto approve because this allows you to preview every action before execution. Nothing happens without your explicit approval."
You can install it very swiftly with just one command. They say it uses claudonnet 4 by default, but you can switch it to GPT5 via an environment variable called copilot model and it should work. Now in the documentation, they do say that GitHub copilot is available with the GitHub Copilot Pro, GitHub Copilot Pro Plus, GitHub Copilot Business, and GitHub Copilot Enterprise plans. But I'm not on any plan and it is working fine for me without any kind of rate limits or restrictions. Probably it's because it is still in public preview. So right now you can use it for free and with Sonnet which is pretty awesome. It works without any rate limits either. I don't know if they'll introduce rate limits later or not, but considering that you'll likely get access to this for just the $10 plan, it's definitely a good deal.
Now, let's get into it and let me show you how it all works. But before proceeding, let me tell you about Ninja Chat. Ninja Chat is an all-in-one AI platform where for just $11 per month, you get access to top AI models like GPT40, Claude for Sonnet, and Gemini 2.5 Pro, all in one place. I've been using Gemini for quick research. But what's really cool is their AI playground where you can compare responses from different models side by side. Their mind map generator is a gamecher for organizing complex ideas as well. The basic plan gives you 1,000 messages, 30 images, and five videos monthly with higher tiers available. If you need more, use my code king 25 for 25% off any plan or king 40 yearly for 40% off annual subscriptions. Check the link in description to try it yourself. Now, back to the video.
So, you install it with the simple command they provide and it installs instantly. To start it, you just run the co-pilot command and it works. If you already have the git CLI, GitHub desktop or copilot in VS Code set up, then it authenticates from there and you can start using it immediately. It's very similar to how cloud code works, looks, and feels. You get some slash commands here like add directory which lets you add only a directory for allowed file access. So if you're just working on the back end, you can simply add the API folder and work with that without touching anything else. You can also run the /mcp command to add MCP servers that you want to use. Generally, I try to add the bite rover MCP and context 7. Byte rover is a memory layer that allows your AI coder to create memories of important things that it may need to remember later and then access them anytime. These memories can also be synced across teams and updated like Git by tracking how an architecture changes over time. You can also manually tweak these memories in many ways, which is pretty cool. You can connect it in Copilot CLI by just following the same steps as Claude Code, and it plugs right in. Anyway, I don't think you can make custom commands or sub agents yet, which I know a lot of people like in Claude Code.
Now, let's try to actually do something with it. I'm going to ask it to make me a movie tracker app using Expo. This is also one of my standard agent testing questions. Once I send it, you'll see that it performs very similarly to Claude. It even uses Sonnet and you'll notice it's basically very similar. However, there are some tools that I found are unique to this like the read bash and write bash tools. I don't think Claude Code has these. What these tools allow it to do is work with interactive shell commands. For example, if it runs a create next app command that requires entering a project name in an interactive shell, then it can do that by using the read bash command and then writing with the write bash command. This is great and actually works very well, which I honestly didn't expect. Anyway, it got the app built, but it gave me one error. I asked it to fix it and then it worked. This is what it looks like. It looks pretty good. There's no padding or margin between the blocks, which is a bit disappointing, but that's more of a clawed issue as it does this very often. Other than that, everything else and the functionality is pretty solid. The git tracker-l like feature also looks nice. So yeah, this is good. Comparing it with Generations from Claude Code, I'd say the one by C-Pilot is a bit better. The one by GLM is also good, but this is one of the most polished and fleshed out ones for sure.
Moving on, I asked it to make me a go terminal. It actually built one, and it looks pretty good. You can see that it works well, which is awesome because Claude Code usually can't get it right on the first or second try, but this one did. The next test was editing an FPS go game to add a step counter and a health bar that changes when you jump. It implemented it pretty well. You can see it works fine. It did take over a large chunk of the UI, which I didn't like. I prefer what DeepSseek does by placing the UI neatly to the right, but still this works and that's great, but I have to say this took way longer than it would with something like DeepSseek or Clawed Code. It constantly wanted to recheck things and at times it tried to fix things that weren't even related to what it wrote. Not the best experience. It does a lot of overesting if that makes sense. Then the last test was adding the SVG generation model in the open code repo and this one failed. It didn't work. So yeah, that was a miss.
Overall, it ranks in third place on my leaderboard. Honestly, that's where I'd put it, too. It's really good, and for now, it's free. I don't have any co-pilot subscription, and it still worked using Sonnet, which is just awesome. I definitely recommend trying it for free while it's available. That's pretty much it. Go ahead and use it all you want. Overall, it's pretty cool. Anyway, share your thoughts below and subscribe to the channel. You can also donate via super thanks option or join the channel as well and get some perks. I'll see you in the next video. Bye. [Music]