Anthropic AI Updates: August 8, 2026

1. Auto Mode Becomes the Default in Claude Code

Anthropic. Anthropic made auto mode the default in Claude Code for Pro, Max, and Team subscribers, so new sessions start with the agent selecting tools and executing multi-step work without the user manually approving each action. A companion post on running auto mode in production offers guidance for teams putting the more autonomous flow in front of real codebases, covering how to scope permissions and where human review still belongs. The change signals Anthropic pushing Claude Code’s default posture from suggestion toward autonomous execution, shifting the developer’s job from driving each step to reviewing outcomes. Source

2. Claude Code v2.1.224 Adds Cross-Session Agent Messaging and Archive Plugin Installs

Anthropic. Claude Code v2.1.224 introduces cross-session SendMessage, letting Claude Code sessions message each other across machines with agent discovery through ListAgents, plus new crossSessionInbound and dialogExpiry settings to manage inter-session delivery. The release also adds an archive plugin source for installing zip-based plugins over HTTPS with optional SHA-256 verification, and sandbox credential-masking options for structured environment values including JWT-aware masking and AWS SigV4 re-signing. Together the additions push Claude Code toward multi-agent coordination and safer handling of secrets inside sandboxed runs. Source