did Anthropic just END OpenClaw?

Quick Overview

Anthropic is not ending support for Claude, but they recently announced changes to their terms of service that effectively wall off their ecosystem, making it difficult or impossible for users to swap out Anthropic's models with open-source alternatives like Claude Code, which was previously used by many developers for custom AI agent loops.

Key Points: Anthropic updated its terms of service recently, which many users interpret as a move against open-source AI agent loops that relied on Anthropic models. The creator of Claude, Peter, had previously spoken positively about the models, including Opus 4.5 and 4.6. Under the new terms, using the API for models like Claude Code is restricted, potentially leading to account bans if users try to use open-source alternatives like Claude Code or other OSS models that leverage Anthropic's infrastructure. The pricing structure for API usage is pay-per-use ($15 per million input tokens, $75 per million output tokens), contrasting with the flat rate of $200/month for the Claude Max subscription which offered unlimited use. The speaker suggests that the core of the issue is that developers who built entire ecosystems around Anthropic's tools are now finding their options restricted, potentially leading to distrust. The speaker notes that the official documentation formalizing these restrictions was published on February 17th, 2026, triggering this shift.

Context: The video discusses recent changes made by Anthropic, the developer of the Claude Large Language Models (LLMs), regarding their terms of service. These changes appear to restrict the ability of users, particularly those building AI agents, to substitute Anthropic's proprietary models with open-source alternatives that might use similar underlying infrastructure or authorization methods, specifically mentioning 'Claude Code' and the concept of 'agentic loops.'

Detailed Analysis

The speaker argues that Anthropic's recent actions, specifically changes to their terms of service announced around January 9th, 2026, essentially end the viability of open-source agentic loops that relied on Anthropic models like Claude Code. Previously, developers could utilize Anthropic's API access, often paired with open-source models, for their custom AI agent loops, which could be cost-effective compared to the API usage fees ($15/$75 per million tokens). However, the new terms explicitly restrict using the authentication key for models other than Anthropic's proprietary ones, effectively walling off their ecosystem. This change means that if a user runs code that attempts to utilize an open-source model like Claude Code (which was previously a popular choice for its unlimited usage under a $200/month subscription), their account risks being banned. The speaker notes that Peter, the creator behind Claude, had previously spoken highly of the models, making this sudden restriction (which they call a 'migration wave' and a 'walled garden') feel like a betrayal to the community members who supported the ecosystem. The speaker concludes that this move damages trust, as users who built entire systems around interoperability now face severe limitations or higher costs, potentially pushing them toward competitors like Google or other open-source providers.