Anthropic is making its CLI and agent workflows more expensive. Starting June 15, 2026, Claude Code and claude -p usage will move out of standard subscription limits into a separate monthly credit pool. Once those credits exhaust, users pay API-style rates. The company also cut off third-party agent frameworks like OpenClaw from subscription access in April, forcing developers to use API billing instead.

This fits a broader pattern. Anthropic announced a $200 million Department of Defense agreement in 2025 and partnered with Palantir for classified networks. The company secured massive compute through a SpaceX partnership giving it access to over 220,000 NVIDIA GPUs at the Colossus 1 data center. Despite positioning itself as a careful, safety-first AI lab, Anthropic hasn't released an open-weight model and keeps advanced capabilities like Claude Mythos behind a partner program.

The company is clearly positioning itself for public markets—tightening monetization, restricting how developers can access its tools, and expanding into enterprise and defense contracts. Each decision makes sense on its own, but together they reveal a strategy focused on controlling access and maximizing revenue rather than maintaining the open, developer-friendly image Anthropic has cultivated.