Anthropic Ends Unlimited Claude Access for AI Agents as White House Showcases Figure 03 Robot

Anthropic ends unlimited Claude access for AI agents to cut costs, while a $25,000 Figure 03 robot is showcased at the White House for education.

By: AXL Media

Published: Apr 7, 2026, 5:23 AM EDT

Source: Information for this report was sourced from CNET

The End of the Unlimited Agentic Era

The landscape of autonomous artificial intelligence is shifting as Anthropic moves to restrict the "all-you-can-eat" model previously available to heavy users of its Claude models. Over the weekend, the firm informed subscribers that utilizing third-party agents like OpenClaw would now require separate billing via an API or a pay-as-you-go option. This policy shift targets the growing tension between users who run autonomous programs constantly and AI labs struggling to manage the immense power and token consumption these agents require. Industry analysts note that the era of inexpensive, constant model access is likely concluding as operational costs continue to climb.

Balancing Systems Strain and Autonomous Demand

Anthropic justified the decision by highlighting the outsized strain that autonomous tools place on its infrastructure compared to standard human-to-bot interactions. While Claude was primarily designed for direct human prompting, agentic tools like Manis and OpenClaw automate complex workflows that burn through compute resources far faster than traditional chatting. To mitigate this impact, the company had previously implemented five-hour session caps during peak periods. The new pricing structure ensures that the costs of high-intensity agent compute are shifted directly onto the power users who utilize them.

Strategic Rivalry and Feature Integration

The policy change has sparked criticism from open-source advocates, including Peter Steinberger, the creator of OpenClaw who was recently hired by OpenAI. Steinberger suggested that Anthropic’s move to lock out open-source tools coincided suspiciously with the company’s internal launch of similar features, such as the ability for Claude to control a computer autonomously. This friction underscores a broader strategic divide between AI labs, with OpenAI appearing to lean into broad agentic access while Anthropic prioritizes a more controlled, managed environment for its enterprise-grade models.