Plans and limits
To build with NodeGX you need a paid Claude plan, because Claude Code is not on the free plan. Pro is enough to start. NodeGX itself is free.
The plans
These prices were checked on claude.com/pricing at the time of writing (September 2026). They change, so check there before you buy.
| Plan | Price | Claude Code? |
|---|---|---|
| Free | $0 | No |
| Pro | $20 a month, or $17 a month paid yearly | Yes |
| Max 5x | $100 a month | Yes |
| Max 20x | $200 a month | Yes |
Team and Enterprise plans include Claude Code too. If your company already pays for one, ask.
Which one to pick
- Free won't work for this. It doesn't include Claude Code, so Claude can't use NodeGX.
- Pro is where to start. You can plan and build a whole first version on it. On a real project you will hit the limits sometimes and have to wait for them to reset.
- Max 5x is what I'd get once you've decided the app is real. The extra room matters most after launch, when you're fixing things and adding features week after week.
- Max 20x is for people who work with Claude all day, every day.
Why you sometimes have to wait
There's no bill per message. Instead, each plan has two limits:
- A session limit, which resets every five hours.
- A weekly limit, which resets at a fixed time each week.
When you reach one, Claude stops until it resets. Nothing is lost. Your app is saved in NodeGX, and you carry on when the limit resets.
Anthropic, the company that makes Claude, doesn't publish exact numbers for these limits, so I won't guess at them here.
What uses your limits up fastest
Long conversations. Claude re-reads the whole conversation every time it answers. A long one costs more on every turn. Start a new conversation for each new job. One job per conversation shows how.
Going round in circles. If the same problem has come back three times, stop. Ask Claude to write down what it has tried, then start a new conversation. When something goes wrong has the words to use.
A big model on a small job. Changing a button's label doesn't need the most powerful model. See below.
If you're close to a limit, don't start something big. Use the time to try what Claude has built, or change things yourself in NodeGX. That doesn't use your Claude limits at all.
Which model to use
Claude comes in several models. A model is one version of the AI. The bigger ones think harder and use more of your limit.
I use Opus for almost everything: planning, building and fixing. If you're not sure, use Opus.
Models in more detail
| Model | What I use it for |
|---|---|
| Opus | Nearly everything. Planning, building, fixing. About 99% of my work. |
| Fable | When the app has a deep problem, or the design isn't working and Opus keeps going round in circles. |
| Sonnet | Small, clearly described changes, especially on Pro when I want my limits to last. |
| Haiku | Quick questions. I rarely build with it. |
At the time of writing, Pro includes Opus, Sonnet and Haiku, with Fable on pay-as-you-go credits. On Max, Fable can use up to half your weekly limit at no extra cost.
On Pro, alternate. Use Opus to plan and for anything tricky. Switch to Sonnet for small, well-described changes. A clear request lets a smaller model do good work.
Switch model with /model. To see how much of your limits you've used, type /usage.
Quick reference
- Paid plan needed. Pro to start, Max 5x once the app is real.
- Two limits: one resets every five hours, one weekly.
- New conversation for each job keeps your limits going further.
- Stuck three times on the same problem? Stop and start fresh.
- Use Opus unless you have a reason not to.