Weekly token limit on Max $200/mo plan after ~1 week on a single solo project — considering canceling

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Other unexpected behavior

What You Asked Claude to Do

I'm on the Max $200/month plan — the top non-enterprise individual tier — and I'm getting "You have used 9X% of your weekly token usage" warnings. I'd like to understand
what's actually happening before deciding whether to continue.

My situation:

• Single developer, working on this in my free time (not a day job)
• Using only Claude Code CLI — no API apps, no other Anthropic tooling
• One project: a simple website, an admin backend, and a data-collection pipeline
• Roughly one week into that project
• No unusual bursts — normal iterative coding, file reads, and agent dispatches

Hitting 9X% of any weekly budget after that amount of work should not be possible on the highest individual tier. If it is, something is off — either the metering itself,
the Max plan's stated value proposition, or my understanding of what the Max plan covers.

What makes this more frustrating:

A non-trivial share of those tokens went into discovering and ticketing issues Claude's agent shouldn't have shipped in the first place — silent error swallowing, React
hydration mismatches, hard-coded timeouts and pagination caps, placeholder data leaking to the UI, half-applied refactors that passed the agent's own self-check and had
to be re-investigated. I'm paying premium pricing for output that then requires premium pricing to audit and clean up. That's tolerable when the pricing is "unmetered,
work freely"; it stops being tolerable the moment a weekly cap enters the picture, because I'm now metered on BOTH the flawed output AND the cleanup of it.

Questions:

1. Is there a weekly token cap on the Max plan? If so, what is it, and is it published anywhere user-visible?
2. Is this warning a hard cap (work stops), a soft cap (throttle), or purely informational?
3. Is there a way to see current usage / budget inside the CLI itself (not just 9X% alarm banners)?
4. Is the Max plan actually intended to support sustained solo-project work at this scope, or is "hit the cap and move to API tier" the expected path?

I signed up for the Max plan specifically because its pitch was "$200/month, stop thinking about tokens." A hidden weekly meter changes that materially, and changes my
calculus about staying on the plan.

Where this lands:

If the Max plan does in fact have a weekly cap and I am actually hitting it with this amount of work, I'll be canceling my subscription and going back to GitHub Copilot
in VS Code. Copilot is about $30/month (nearly 10× less). It's not as capable, it needs more hand-holding, and I'll be doing more of the work myself — but it doesn't
rate-limit me after a week of moderate solo use, and I won't be cleaning up mediocre output against a meter.

I'd rather not cancel. The tool is genuinely capable when it works. Please clarify whether this warning reflects a real cap, and if so, whether the Max plan is the wrong
tier for this use case.

If this is the wrong place to put this, sorry please redirect it, Claude Code AI Agent is what told me it belongs here.

What Claude Actually Did

Used up my tokens.

Expected Behavior

Not use them at such a high rate.

Files Affected

all of them in my project

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

use Claude Code on a single small to medium sized single "Hobby" developer project.

Claude Model

Sonnet

Relevant Conversation

I'm just a guy on a tight "Out of my own pocket" budget doing this in my free time.

Impact

Critical - Data loss or corrupted project

Claude Code Version

2.1.100

Platform

Anthropic API

Additional Context

_No response_