Anthropic Under Fire for Egregious Billing Practices and Anti-Developer Stance

Anthropic’s Claude Code service is facing intense scrutiny following multiple reports of contentious billing practices and anti-third-party tool measures. One particularly egregious incident involved a user on a $200/month Claude Code subscription who was reportedly billed the full amount despite having significant unused usage. The charge was triggered not by actual model inference, but by the inclusion of “Hermes MD” within a git commit message, which was then inadvertently passed into the Claude Code system prompt through its programmatic claude -p interface. This unexpected charge occurred because Anthropic’s internal “third-party harness detection” mechanisms—intended to prevent use with tools like OpenClaw or Hermes agent—flagged the string, rerouting the user to an overage billing system even if their subscription limits hadn’t been met. This incident underscores broader concerns about Anthropic’s attempts to lock users into its proprietary Claude Code harness, often by penalizing the use of external tools or specific system prompt structures.

Beyond these billing anomalies, Anthropic’s technical infrastructure and developer relations are also drawing criticism. Reports indicate severe inefficiencies in Anthropic’s caching system, which paradoxically charges users for cache writes (a practice not observed in competitors like OpenAI and Google) and has seen cache persistence reduced from an hour to mere minutes. A large-scale user, T3 Chat, reported that cache writes accounted for approximately half of their $40,000 monthly Anthropic bill, yet disabling caching entirely yielded no meaningful cost increase, suggesting the system’s ineffectiveness. Furthermore, Anthropic’s API service has been cited for unreliability, leading to instances where automatic billing failed, causing service disruptions for downstream users and erroneous non-payment notifications. These technical issues, combined with a perceived lack of transparency and an alleged disdain for engineers from Anthropic’s leadership, are eroding trust within the developer community. Critics argue that this approach negatively impacts the overall developer experience and places Anthropic at a disadvantage compared to competitors like OpenAI, which are lauded for more stable services, competitive pricing, and proactive engagement with user feedback.