Tell HN: Bug in Claude Code CLI is instantly draining usage plan quotas
agents anthropic claude
| Source: HN | Original article
Anthropic’s Claude Code command‑line interface is suddenly exhausting user quotas at an alarming rate, a problem first flagged by developers on the “Tell HN” forum over the weekend. According to a GitHub issue, premium plans that normally last weeks are being drained to 100 % in ten to fifteen minutes, even when the tool reports cache‑hit rates above 98 %. The CLI appears to hit rate limits on every request, inflating usage counters regardless of whether the underlying model call is served from cache.
The glitch matters because Claude Code is a cornerstone of Anthropic’s developer offering, bundled with Team and Claude Max plans and marketed as a drop‑in alternative to OpenAI’s Codex. Its promise of self‑serve seat management and “extra usage at standard API rates” has attracted enterprises that rely on the tool for automated file editing, code generation and other agentic tasks. Rapid quota depletion not only spikes costs for customers but also erodes confidence in Anthropic’s billing transparency—a concern already highlighted in our March 30 AI‑rationing piece on Claude Code promotions.
Anthropic has not yet issued an official statement, but the company’s engineering team is reportedly investigating whether the problem stems from a mis‑counted cache‑hit metric or a deeper fault in the CLI’s rate‑limit logic. Users are advised to monitor the “usage counter” in their Claude Max sessions and consider throttling calls until a fix lands.
What to watch next: a patch or rollback of the usage accounting, potential compensation for affected accounts, and any changes to the CLI’s caching strategy. The incident also raises the question of whether similar bugs could surface in related tools such as the Agentic Shell layer we covered earlier. Developers will be keeping a close eye on Anthropic’s response, as the resolution will influence whether Claude Code remains a viable component of Nordic AI‑driven development pipelines.
Sources
Back to AIPULSEN