2026-05-21 · Anthropic

v2.1.146

protocolsagentsmodels

read at source ↗ github.com

v2.1.146

Source: Anthropic Claude Code Date: 2026-05-21 URL: https://github.com/anthropics/claude-code/releases/tag/v2.1.146

Summary

Claude Code v2.1.146 (released 2026-05-21) is a platform-stability and cross-platform-parity release. The headline change is renaming /simplify to /code-review with an optional effort level (/code-review high). Beyond that, the release closes a Windows regression from v2.1.124 where PowerShell failed with winget/Store installs, fixes MCP pagination (items past page 1 were silently dropped on paginating servers), and resolves several multi-agent issues: CLAUDE_CODE_SUBAGENT_MODEL not forwarding to child processes, and backgrounded sessions re-prompting for already-granted tool permissions. Enterprise policy enforcement for forceLoginOrgUUID and forceLoginMethod was broken against third-party-provider sessions — now fixed.

Implications

Feeds the coding-agent landscape and agentic orchestration patterns threads:

  • The /code-review rename with effort levels is a UX signal that Anthropic is formalizing code review as a first-class workflow, not just a slash-command alias
  • MCP pagination fix is load-bearing for any server exposing large resource lists; silent truncation was a latent data-loss bug for orchestration setups
  • The CLAUDE_CODE_SUBAGENT_MODEL forwarding fix matters for multi-agent pipelines that route different sub-tasks to different model tiers — the regression broke model routing in child processes
  • Enterprise policy enforcement fixes tighten the institutional deployment story ahead of broader enterprise rollout

← all signals