Switching from Claude Code to Grok – Same Interface, Different Model – Dragos Roua

At the beginning of June I started a “Claude withdrawal” challenge. The plan was to run MiniMax 3 for a month, to see if I can get the same level of quality, but at 5x less the price. Until then, Claude Code was my main driver, with MiniMax on the backup, for when I was running out of quota, or sometimes for code review. The monthly bill for Claude was $100 on the Max plan, whereas for MiniMax I would pay $20 for the Token plan. All in all, it seemed like an interesting experiment.

Then, half way through the challenge, Grok came into the picture. I got a very interesting offer at $35 for 3 months, then $35/month. But Grok has something neither Claude, nor MiniMax can give me out of the shelf: video and image generations. The only unknown was if switching from Claude Code to Grok will still maintain the same coding power. So I instantly took the offer, and did whatever I had to do to understand if this was the right path.

And here comes the “whatever I had to do”, in plain technical terms.

Switching from Claude Code to Grok – the Actual Steps

The switch itself was interesting because I didn’t want to lose the Claude Code interface. I like the harness. The way it works with my codebase, the commands, the flow. So I used a helper called cliproxyapi. It’s a small proxy that sits between the Claude Code client and whatever model you point it at. You run it locally, tell it to forward requests to Grok’s API instead of Anthropic’s. Then you launch Claude Code the same way you always do, but it talks to Grok under the hood.

Here’s how it goes in practice.

Step 1: Install the proxy. I used brew to install it, I’m on a Mac, and also because I wanted to have it started as a service.

Step 2: Set two environment variables. One is the target API base URL, for Grok that’s something like https://api.x.ai. The other is your API key.

"env": {
    "ANTHROPIC_BASE_URL": "http://localhost:8317",
    "ANTHROPIC_API_KEY": "cliproxy-local-key"
  },

Notice how we use “cliproxy-local-key”, because, and here is where it gets interesting, you log in with xAi OAuth, so you just stay inside Grok standard package, not the one using API tokens. It’s the same with Claude Max, I always logged in with OAuth and kept my usage limits inside the Max plan.

Step 3: Because you have to be logged in via OAuth to xAI, you just use this command before starting cliproxyapi:

cliproxyapi -xai-login

After that you just run claude like normal. The proxy intercepts the calls, rewrites the auth headers, and sends everything to Grok. It handles the main chat loop and the tool use stuff too. I’m running it for about a week without any hiccups.

Is It Worth Switching from Claude Code to Grok?

What I get out of this is the same coding experience I had with Claude, maybe better on some things, plus full access to Grok’s image and video generation. And I’m paying $35 instead of $100. Not even that, because the $35 plan will kick in about two and a half months. I already cancelled my Claude and MiniMax subscriptions.

And then there’s this: Grok is less censored. Claude has a lot of guardrails. Some of them make sense, some of them are just annoying when you’re trying to explore an idea or write something that touches on certain topics. Ask it to roleplay something edgy or dig into a controversial angle and it starts hedging or refusing outright. Grok lets you push further. For the kind of work I do, where I’m often asking for unfiltered takes or edge case scenarios, that matters. I don’t want to fight the model just to get it to consider an uncomfortable premise.

The prices in AI keep moving. Five months ago I was at $100. Beginning of June I was at $120 (Claude + MiniMax). Now I’m at $35 but with capabilities I didn’t have in either of the previous setups. The market is still figuring itself out. Providers are competing hard on price because the models are getting better fast and nobody has a permanent moat yet. That works in the user’s favor for now.

But the end goal isn’t to find the cheapest subscription that does everything. The end goal is to run this stuff locally. I’ve already written about running private, local models before. When your tools can be taken away by a policy change, or a regional restriction, or just a company deciding you’re not worth serving anymore, you’re not in control. Local models don’t have that problem. No rate limits. No one watching what you generate. No sudden deprecations.

The gap is closing. Every time the hosted prices drop and the quality goes up, it gets easier to see the path to running something comparable on your own hardware. Not there yet for video generation at the level Grok Imagine does. But the coding side is already viable with some of the open source models, if you’re willing to put in the setup work.

A year ago the gap felt huge. Now it feels like a matter of time and hardware budget.

This June challenge started as a test of MiniMax. It turned into a test of how I actually want to work with these tools. The answer so far is: keep the harness I like, add the capabilities I need, and don’t get too attached to any single provider.