Claude Code Is Burning Through Your Token Budget—And Anthropic Knows It

If you've been using Claude Code and wondering why your daily token limit hit 100% faster than you expected, you're not alone. Users are reporting that they're hitting usage caps "way faster than expected," and Anthropic has acknowledged the issue as their "top priority."

What's Happening

The problem seems to be that Claude Code is consuming tokens at an unexpectedly high rate. Users report hitting their daily limits in hours or even minutes when running code loops or intensive tasks.

One Reddit user noted they hit their token limit "much later" on their free account compared to their $100/month paid account. Another reported that "one session in a loop can drain your daily budget in minutes." And someone else said a single one-sentence reply to a conversation jumped their usage from 59% to 100%.

For context: Claude Pro costs $20/month. Higher tiers go up to $100-200/month. For developers who use Claude Code as part of their daily workflow, these kinds of overage issues aren't just annoying—they're disruptive.

Why It Matters

Claude Code is designed to help software developers with specific coding tasks. It should be a reliable tool in their daily workflow, not a budget-draining surprise.

When the service eats through your allocation faster than you expect, it breaks workflow. Developers have to either pay more, stop using the tool, or plan around unpredictable token consumption.

The Context

This comes on the heels of several other Claude-related issues:

Peak-hour throttling: Last week, Anthropic introduced peak-hour throttling, meaning tokens get consumed faster when demand is higher. That's a reasonable load-balancing approach, but it compounds the unpredictability.
Accidental source code leak: Earlier this month, Anthropic accidentally released 500,000 lines of Claude Code's internal source code on GitHub due to "human error." The company says no customer data or credentials were exposed.
Earlier leaks: Claude Code's source had already been partially reverse-engineered by independent developers, with earlier versions leaking in February 2025.

What's Next

Anthropic says fixing the token consumption issue is their top priority. But "top priority" is developer-speak for "we know about it and we're working on it"—which doesn't help if you just burned through your monthly budget on a coding session.

For now, if you're a Claude Code user, the advice is simple: monitor your token usage closely, use the free tier carefully to test workflow, and be prepared for surprises. And if you're considering paying for a higher tier, know that Anthropic is aware the pricing-to-performance ratio is currently broken.

This is the kind of issue that separates good AI tools from great ones. Anthropic built something genuinely useful. But usefulness breaks down when the cost structure is unpredictable.

Source: BBC News - Claude Code users hitting usage limits