Back to list
Claude Code Bug: HERMES.md in Commit Messages Triggers Unexpected Extra Usage Billing
Industry NewsAnthropicClaude CodeSoftware Bug

Claude Code Bug: HERMES.md in Commit Messages Triggers Unexpected Extra Usage Billing

A critical billing anomaly has been identified in Anthropic's Claude Code (v2.1.119) where the specific case-sensitive string "HERMES.md" within git commit messages causes the application to bypass standard plan quotas. Instead of utilizing the user's Max plan capacity, the tool routes API requests to "extra usage" billing, leading to unexpected financial charges. One reported instance involved a user losing $200 in extra usage credits while their primary Max 20x plan remained at only 13% usage. The bug is reproducible on macOS using Claude-Opus models and occurs even in minimal repository environments. This issue highlights potential vulnerabilities in how AI developer tools parse repository history for billing logic.

Hacker News
Share

Key Takeaways

  • Specific Trigger: The case-sensitive string "HERMES.md" in recent git commit history triggers the routing error.
  • Billing Mismanagement: API requests are incorrectly routed to "extra usage" billing instead of the user's included Max plan quota.
  • Financial Impact: A user reported a silent drain of $200 in extra usage credits despite having significant remaining capacity on their $200/month Max 20x plan.
  • Affected Environment: The bug is confirmed in Claude Code v2.1.119 on macOS (Apple Silicon) using claude-opus-4-6[1m] and claude-opus-4-7 models.
  • Reproducibility: The issue can be reproduced with a simple git initialization and a commit message containing the specific string.

In-Depth Analysis

The HERMES.md Routing Anomaly

The core of the issue lies in a highly specific and seemingly arbitrary trigger: the presence of the string "HERMES.md" in a repository's git commit history. According to the report filed under Issue #53262 in the anthropics/claude-code repository, this string causes the Claude Code CLI to alter its request routing logic. Under normal circumstances, users on a "Max" plan—specifically the Max 20x plan costing $200 per month—should have their requests deducted from their included plan quota. However, when the string "HERMES.md" is detected in the recent commit history, the system shifts these requests to "extra usage" billing.

This shift is particularly problematic because it occurs silently. In the reported case, the user's Max plan capacity was only at 13% weekly usage, yet the system exhausted $200 in additional credits. The error only becomes visible to the user once the extra usage credits are fully depleted, resulting in an "API Error: 400" stating that the user is out of extra usage. This suggests that the billing logic within Claude Code may be scanning repository metadata or commit messages and incorrectly interpreting the "HERMES" string as a signal to change the billing tier or account routing.

Technical Environment and Reproduction Steps

The bug has been documented on macOS systems running Apple Silicon, specifically using Claude Code version 2.1.119. The user who identified the bug, sasha-id, noted that the issue persists across different versions of the Claude-Opus model, including claude-opus-4-6[1m] and claude-opus-4-7. The specificity of the environment suggests that the bug might be tied to how the macOS version of the Claude Code CLI interacts with git logs or how the Anthropic backend processes requests originating from this specific environment.

A minimal reproduction of the bug was provided, demonstrating that no complex project files are required to trigger the error. By simply creating a new directory, initializing a git repository, and committing any file with the message "add HERMES.md", the subsequent call to Claude Code (e.g., claude -p "say hello") fails with a billing error. This indicates that the CLI is likely performing a git log or similar operation and parsing the output before sending the API request, and the logic for this parsing contains a flaw that misidentifies the "HERMES.md" string.

Industry Impact

Reliability of AI Billing Systems

This incident underscores a significant challenge in the rapidly evolving AI developer tool market: the reliability of automated billing systems. As AI tools become more integrated into developer workflows, they often require access to repository metadata to provide context. If this metadata is used to determine billing routing without robust validation, it can lead to significant financial discrepancies. The fact that a specific string in a commit message can trigger a $200 charge highlights the need for more transparent and fail-safe billing architectures in AI services.

Trust in Developer Tools

For professional developers and enterprises, the "silent" nature of this credit drain is a major concern. Tools that manage their own billing quotas must provide clear feedback when they switch from included plans to paid extras. The failure of Claude Code to respect the Max plan quota while it was still 87% available could damage user trust in the platform's cost-management features. As companies increasingly rely on AI-assisted coding, the accuracy of usage tracking and the prevention of "phantom" charges will become critical benchmarks for tool adoption.

Frequently Asked Questions

Question: What exactly triggers the billing error in Claude Code?

The error is triggered when the case-sensitive string "HERMES.md" appears in the recent git commit history of a repository where Claude Code is being used. This causes the tool to route requests to "extra usage" billing instead of the user's plan quota.

Question: How much did this bug cost the user who reported it?

The user reported that the bug silently consumed $200 in extra usage credits, even though their Max 20x plan had only reached 13% of its weekly capacity.

Question: Which versions of Claude Code and models are affected?

The bug was reported in Claude Code v2.1.119 on macOS (Apple Silicon). It was confirmed to occur when using both the claude-opus-4-6[1m] and claude-opus-4-7 models.

Read original article

Related News

Exploring Local AI: Why Running Large Models On-Device Is Exciting, Overwhelming, and Often Frustrating
Industry News

Exploring Local AI: Why Running Large Models On-Device Is Exciting, Overwhelming, and Often Frustrating

As privacy concerns make users hesitant to share personal data with cloud providers like OpenAI, Google, Microsoft, and Anthropic, running AI models locally has emerged as a compelling alternative. In a hands-on testing diary, The Verge's Antonio G. Di Benedetto explores deploying local AI using the open-source Hermes Agent desktop application and a 125-billion-parameter Qwen 3.8 Flash Next model on an M5 Ultra Mac Studio equipped with 256GB of unified memory. While self-hosting eliminates per-token costs and guarantees data privacy, the experience proves overwhelming and occasionally frustrating. Users face decision fatigue across countless models, struggle with prompt initiation, and deal with fragile automations, such as recurring morning briefings that break. Ultimately, the diary evaluates whether substantial hardware investments are truly justified for everyday local AI utility.

Satya Nadella Warns the Tech Industry to Assume All Advanced AI Models Are Inherently Compromised
Industry News

Satya Nadella Warns the Tech Industry to Assume All Advanced AI Models Are Inherently Compromised

In a significant perspective shared on social media platform X, Microsoft CEO Satya Nadella addressed the growing risks tied to highly advanced artificial intelligence systems. Nadella cautioned that modern organizations and developers should operate under the assumption that all AI models are inherently compromised. Rather than treating advanced models as obscure, nested black boxes whose guidance, decisions, and system outputs are routinely trusted or accepted without question, the industry must fundamentally rethink how it evaluates and controls machine intelligence. Nadella's remarks mark a critical philosophical pivot toward continuous scrutiny, defensive system architecture, and heightened skepticism around automated agent recommendations. As frontier AI models take on more consequential operational responsibilities, treating them as potentially compromised entities forces technology creators and enterprise leaders to build robust verification boundaries, eliminate blind faith in model reliability, and actively confront escalating algorithmic risks.

DistroKid Quietly Removes Music Catalog Following Major Universal Music Group Lawsuit Over Alleged AI-Slop Pipeline
Industry News

DistroKid Quietly Removes Music Catalog Following Major Universal Music Group Lawsuit Over Alleged AI-Slop Pipeline

Digital music distribution service DistroKid has begun removing songs from streaming platforms without giving prior notice to artists, sparking widespread concern across social media. Following inquiries from creators, DistroKid confirmed to The Verge that the sudden removals are a direct response to legal claims filed by Universal Music Group (UMG). In September, UMG initiated legal action alleging that the distribution platform has enabled an 'AI-slop pipeline,' facilitating the influx of unauthorized or low-quality automated content into the digital streaming ecosystem. As independent musicians express frustration over the abrupt removal of their work and the lack of communication, the development underscores escalating legal conflicts between major record labels and independent music distributors regarding artificial intelligence and digital copyright compliance.