[BUG] Billing : Had a Monthly Limit set to 50$ still it didn't stopped there and consumed more.

Found 3 possible duplicate issues:

1. https://github.com/anthropics/claude-code/issues/1047
2. https://github.com/anthropics/claude-code/issues/14229
3. https://github.com/anthropics/claude-code/issues/17694

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

Additional Report — Same bug, extra details

I experienced the same issue. Set a R$10 (BRL) monthly spending limit as a test, and was charged R$27.36 (274% of the limit) in a single session using Claude Code CLI with Opus (1M token context).

Key additional observations:

1. No progressive billing feedback during usage: Unlike other models that show billing progression in real-time, the Opus 1M context session charged everything all at once after the session ended — no intermediate warnings that the limit was being approached or exceeded.

1. No session interruption: The system should have stopped or warned when approaching the R$10 limit, but instead allowed the full R$27.36 charge to go through.

1. Environment: Claude Code CLI, Opus model, 1M token context window, auto-reload disabled.

This suggests the spending limit enforcement is either not checked during long-context sessions, or the billing granularity is too coarse (single large charge instead of incremental).

Originally reported in #27154

Additional feedback — Extended context (1M tokens) billing expectations

Following up on my previous comment: there's also a billing model clarity issue with the extended 1M token context.

The expectation: If the plan (Max, Pro, etc.) advertises "extended context up to 1M tokens" as a feature, users naturally expect that usage to be covered by the monthly plan quota — not billed separately as extra API usage.

What actually happens: Using the 1M token context in Claude Code silently consumes API credits (extra usage) rather than the plan's monthly allowance. There's no clear indication that switching to extended context changes the billing model from "plan quota" to "pay-per-token API credits."

Why this matters:

• Users who received promotional API credits (like myself, $275 to test) expect them to last a reasonable amount of time
• The cost per interaction grows exponentially as context accumulates (each message re-sends the full context)
• A few messages in a single session can cost $25+ without any warning
• Combined with the spending limit enforcement bug, this creates a poor billing experience

Suggestion: Either include extended context usage within the plan quota, or make it very explicit when the billing model switches from plan-included to API-credit consumption — ideally with real-time cost feedback during the session.

ngl sounds like the $50 cap didnt actually block some requests so charges kept piling up, check ur dashboard for pending usage and file a support ticket if it keeps billing. ps wozcode cut my token spend about 50% with smarter caching, worth trying https://wozcode.com