robeym

I’m trying to understand what the alternative workflow looks like if you use a Claude subscription instead of API billing.

Right now I use Claude Code in terminal with an API key and spend about $150–300/month.

My February usage looked like:

- Haiku: 300M tokens in, 2M out (~$60)

- Sonnet 4.5: 100M in, 1M out (~$75)

- Sonnet 4.6: 20M in, 300k out (~$20)

Most of my work is with Haiku. I mainly use it for simple tasks such as small customizations for a customer or basic feature additions.

My workflow is simple. I stay in the terminal and pass context with things like `@file1` and `@file2`. For database work/references I query things in pgAdmin and paste output into the session.

I’ve had people say the Pro/Max subscriptions are cheaper and still include Claude Code, but I don’t understand what actually changes in practice. Claude Code can authenticate either with a Claude account or an API key, and if `ANTHROPIC_API_KEY` is set it will bill the API instead of the subscription, I believe.

If you use the subscription with Claude Code:

- Is the terminal workflow basically identical?

- Do you still just run `claude` and use `@files` the same way?

- Do you run into any annoying limits (context or monthly)?

Trying to understand whether switching from API to a subscription would change anything for someone who already works in the terminal. I'm also curious if I could have greater stronger model usage?

I'm a developer who built an ERP/CRM system for small manufacturers (https://www.paxerp.com). It does all the basics very well; financial reporting, lot tracking, production planning, shipping carrier integrations, the usual workflow stuff—but there's zero AI in it. It's just fast, clean, and solves real problems I saw working in manufacturing ERP systems. The product works well, customers really like it, but I have almost no sales experience.

Every SaaS founder seems to be talking about "AI-powered insights" and "intelligent automation." while... We just have a clean system that is fast and tries to stay out of the user's way.

For those who've sold B2B SaaS (especially to traditional industries like manufacturing):

- Is "no AI" actually a disadvantage, or does it not matter as much as I think?

- How do I communicate value when the value is "it's simple and fast, and your data is highly accessible" vs "revolutionary AI"?

- Should I be adding AI features just to check a marketing box, even if customers don't need them?

You can learn more about why I built this on the websites about page. But now I'm wondering if I'm fighting an uphill battle by not having the buzzwords everyone else does.

Any advice from founders who've been here?

TY