Anthropic is facing intensifying criticism from the developer community regarding a series of policy changes that limit subscription flexibility, increase costs for third-party agent usage, and tighten vendor lock-in. Industry observers highlight that these moves, combined with reported API instability and software quality issues within the Claude Code environment, are eroding trust and accelerating a migration toward open-source alternatives. A pivotal change took effect on June 15, introducing a split in billing structures that fundamentally alters the value proposition for heavy users. Anthropic now separates usage between first-party applications and third-party agent integrations, including those utilizing the Anthropic Client Protocol or command-line invocations. Subscriptions no longer subsidize third-party agent calls; instead, each tier receives a limited monthly Agent SDK credit priced at standard API rates. Pro users receive $20 in credits, while Max tier plans offer between $100 and $200. Once these credits are exhausted, continued usage triggers standard per-token billing or halts requests entirely. This shift effectively nullifies the previous subsidy, where subscription users enjoyed agent pricing roughly 15 to 30 times lower than direct API access, resulting in significant cost increases for workflows relying on external tooling. The new restrictions reinforce Anthropic's closed ecosystem strategy. The Claude subscription is now functionally restricted to Anthropic's native interfaces, such as the web chat, Claude Code, and Slack integration. Attempts to leverage subscriptions through third-party harnesses have been met with retroactive billing adjustments, drawing from extra usage pools that incur API rates regardless of the user's plan. This policy has been accompanied by technical grievances. The Claude API remains prone to reliability issues, and the Claude Code interface has accumulated thousands of open GitHub issues, including persistent bugs related to freezing and display artifacts. Critics argue that the focus on locking users into a proprietary stack detracts from software stability and innovation. In response to these constraints, developers are increasingly exploring open-source models and flexible AI gateways. Models such as Qwen and GLM have demonstrated competitive performance in coding benchmarks, offering high-quality results without the prohibitive costs of proprietary subscriptions. Gateways enable users to route requests across multiple backend providers, optimizing for cost, availability, and data privacy. The backlash extends beyond pricing; industry commentary suggests that the push toward agent-driven workflows may undermine software engineering discipline, with many professionals reverting to agent-assisted workflows that prioritize human oversight. The cumulative effect of these developments has sparked a broader conversation about ethical AI practices. Developers cite artificial restriction, dynamic pricing experiments, and post-sale term changes as behaviors that alienate the professional community. As proprietary solutions become more restrictive, momentum is shifting toward modular, open architectures that promise greater control, transparency, and long-term sustainability for engineering teams.