Descriptions:
Matthew Berman reacts to Anthropic’s decision to ban third-party harnesses from Claude Pro subscriptions — explicitly calling out OpenClaw by name — effective April 4th with less than 24 hours of notice. The video covers what drove the policy change, what it means for developers currently using agentic workflows through Claude, and how quickly the switch to alternative models can actually be made.
The business logic is straightforward once laid out: Cursor’s data suggests a $200 Claude subscription translates to roughly $2,000 worth of API credits, meaning heavy users running agentic workflows through OpenClaw are consuming far more than their subscription price covers. Anthropic has been managing this through a combination of incentives — doubling usage limits during off-peak hours (weekdays outside 5–11 AM Pacific, and all day weekends) — and enforcement, including accelerated session limit consumption during peak hours for the top 7% of users. Banning OpenClaw outright is the most recent and most direct measure.
Berman demonstrates that switching models is frictionless in practice: he migrated from Claude to GPT-5.4 through OpenClaw in roughly three minutes using pre-optimized prompt files. He also highlights a live policy ambiguity that remained unresolved days after the announcement — whether the Anthropic Agents SDK still qualifies under the new rules — and notes that OpenAI’s comparatively liberal quota policies make it an easy default fallback. The episode serves as both a practical guide for affected users and a broader look at the capacity constraints reshaping how frontier AI subscriptions work.
📺 Source: Matthew Berman · Published April 08, 2026
🏷️ Format: News Analysis
