Anthropic AI Updates: May 12, 2026

1. Agent view lands in Claude Code as a unified session dashboard

Anthropic. Anthropic shipped agent view, a centralized management interface in Claude Code that lists every session — running, blocked, or completed — and lets developers launch multiple agents in parallel, peek at responses, reply inline, and background sessions without leaving the dashboard. The feature ships as a Research Preview in v2.1.139 and is accessible via the claude agents command. Source

2. Claude Platform on AWS reaches general availability

Anthropic. Anthropic launched the Claude Platform on AWS for general availability, giving AWS customers full access to the native Claude API — including Claude Managed Agents, code execution, web search, skills, and prompt caching — through AWS authentication, billing, and IAM. The offering sits alongside Claude on Amazon Bedrock and routes data processing through Anthropic rather than the AWS boundary, letting organizations pick between Anthropic-operated infrastructure and AWS-managed data handling. Source

3. Claude Code v2.1.139 adds /goal, hook exec form, and richer plugin tooling

Anthropic. Claude Code v2.1.139 introduces a /goal command that sets a completion condition and keeps Claude working across turns in interactive, -p, and Remote Control modes with a live elapsed/turns/tokens overlay. The release also adds a /scroll-speed command, transcript view keyboard navigation (?, {/}, v), a claude plugin details <name> inventory and per-session token cost view, a new hook args: string[] exec form that skips the shell, a continueOnBlock option for PostToolUse hooks, and CLAUDE_PROJECT_DIR in MCP stdio server environments. Source