The hard part of building a free Claude Code alternative isn't the agent. It's paying for the tokens.

I've been looking at a lot of free AI coding agents recently, and I noticed that "free" usually means one of a few things:

You run a local model.

You get a small free tier.

You bring your own API key.

Or the tool itself is free, but the AI usage isn't.

I wanted to try something different.

So I'm building a free AI coding agent for the terminal that works more like Claude Code or Codex CLI, but without requiring users to bring their own API key or pay for a monthly subscription.

The agent can read and search your project, edit files, run commands and tests, and show you a diff before making changes.

The difficult part was figuring out how to actually pay for the inference.

The model I'm testing is pretty simple:

You get free credits when you sign up.

When you run out, you can complete a form or survey from an advertiser. The advertiser pays for the completion, and most of that money gets converted into AI credits.

So instead of paying with a credit card, you pay with a few minutes of your time.

It's obviously not for everyone. If you're using an AI coding agent all day, paying for a subscription is probably a much better deal.

But for students, indie hackers, hobby developers, or anyone who only needs a coding agent occasionally, I think there might be something here.

The question I'm trying to answer is:

Would you actually use a free Claude Code alternative or free Codex CLI alternative if the tradeoff was occasionally completing a short offer instead of paying a subscription?

That's the part I'm still trying to figure out.

I'm building it as Clixad and the waitlist is open.