Claude Code Allegedly Penalizes Users for Mentioning Competitors

Theo revealed that Claude Code penalizes users who have 'OpenClaw' mentioned in their repository's commit history or JSON blobs. This penalty includes refusing requests or charging extra money, even in empty repositories. The developer community has reacted with outrage, calling the move a desperate attempt at gatekeeping that ruins Anthropic's public image.

Key Points

• Claude Code allegedly detects mentions of the competitor 'OpenClaw' in repository commit history and JSON files.
• Users found that mentioning this keyword results in service refusal or extra charges.
• The behavior persists even in empty repositories, indicating a direct scan of the environment for specific strings.
• The developer community views this as a breach of trust and an anti-competitive move by Anthropic.
• Critics argue that Anthropic is damaging its public image by using 'regex duct tape' to gatekeep its ecosystem.

Sentiment

The overwhelming sentiment is strongly negative toward Anthropic. The community views this as either deliberate anti-competitive behavior or negligence so severe it amounts to the same thing. Even commenters who acknowledged Anthropic's economic challenges with subscription pricing condemned the crude, opaque implementation. The incident reinforced a growing narrative that Anthropic's stated values of safety and user-friendliness are increasingly at odds with its actual business practices.

In Agreement

• Multiple users independently reproduced the issue, confirming that a commit message containing OpenClaw metadata triggers session disconnection and quota consumption in Claude Code
• The detection system is weaponizable — anyone can embed OpenClaw strings in open-source repos, blog posts, or documentation to sabotage Claude Code users who interact with that content
• This fits a pattern of anti-consumer behavior at Anthropic, including expiring prepaid API credits without refund and only issuing refunds after issues go viral publicly
• The implementation was likely vibe-coded with minimal human review, reflecting dangerous incompetence at a company that claims to be building safe AI
• Silently draining user quotas or charging extra fees based on keyword detection without disclosure or consent is potentially illegal and certainly unethical
• Anthropic's public posture of building AI 'for the benefit of humanity' is incompatible with anti-competitive keyword blocking in paid products

Opposed

• Subscription users consuming far more in API tokens than they pay creates an unsustainable economic model, making some form of usage management inevitable
• Hanlon's razor may apply — this could be a bug in usage segmentation rather than deliberate malice
• Some users could not reproduce the issue, and one person reported using Claude Code with an actual OpenClaw instance for weeks without problems
• If OpenClaw is specifically designed to exploit Claude's subscription pricing by running agents in loops, blocking it may be economically justified even if poorly executed
• Anthropic reportedly offered refunds to affected users after the issue became public