Anthropic News Flash | Claude Code v2.1.146 and /code-review
The Clauder Navi editorial team brings you the official Anthropic updates for May 22, 2026 (JST) in a news broadcast style. Today's headlines cover two stories: the release of Claude Code v2.1.146, which renames /simplify to /code-review, and the Day 1–Day 2 event report from Code w/ Claude London, as covered on-site by Fortune and MIT Technology Review.
Today's most important topic is Claude Code v2.1.146, which brings a significant overhaul to review-related commands. /simplify has been renamed to /code-review, and you can now specify an effort level as an argument — for example, /code-review high. If any integration scripts or documentation reference /simplify, an immediate name replacement is required. Additionally, auto mode has been updated so that AskUserQuestion is no longer suppressed when the user or a skill explicitly depends on it, ensuring that necessary confirmations are not dropped during automated runs.
On the security and stability front, a "command line is invalid" regression that had affected winget / Microsoft Store installations of Claude Code on Windows PowerShell since v2.1.124 has been resolved. A pagination bug causing pages 2 and beyond to be dropped in MCP's resources/list, resources/templates/list, and prompts/list has also been fixed. The issue where forceLoginOrgUUID / forceLoginMethod policies were not applied to third-party provider / API key sessions has been resolved as well, finally making enterprise access controls effective. Several other issues impacting Linux and Windows professional users have also been addressed, including GNOME Terminal right-click/middle-click paste, propagation of CLAUDE_CODE_SUBAGENT_MODEL to multi-agent child processes, and a full-screen strobe effect in Windows Terminal background sessions.
The second major story is the Code w/ Claude London Day 1–Day 2 event report. Fortune and MIT Technology Review published on-site coverage on May 21, 2026, reporting that the four items from the May 19 release notes (MCP Tunnels research preview, Managed Agents self-hosted Sandbox, live MCP server configuration updates, and automatic spilling for outputs exceeding 100K tokens) were explained in detail during the Day 1 keynote. Customer case study sessions from Asana, Cursor, GitHub, Replit, and Vercel were also held, and Claude for Small Business (released May 13) was re-introduced at the venue.
目次 (4)
- [1] Claude Code v2.1.146 Released — /code-review Rename and Auto Mode Confirmation Bug Fix
- [2] Code w/ Claude London Day 1–Day 2 Event Report — On-Site Coverage by Fortune and MIT Tech Review
- [What to Watch Next] Immediate Adoption of v2.1.146 and /code-review Name Replacement
- [Sources] Primary Source Links Referenced in This Article
[1] Claude Code v2.1.146 Released — /code-review Rename and Auto Mode Confirmation Bug Fix
We're reporting that Claude Code v2.1.146 was released on May 21, 2026 at 01:51 UTC (primary sources: https://github.com/anthropics/claude-code/releases/tag/v2.1.146 and https://code.claude.com/docs/en/changelog). This release brings an overhaul of review-related commands alongside numerous stability fixes.
Impact for engineers: ★★★
The biggest change is the renaming of /simplify to /code-review. Effort levels can now be specified as arguments (e.g., /code-review high), allowing review depth to be tuned based on the scope of changes. Since the /simplify name has been retired, any references in CI scripts, documentation, or team runbooks must be updated immediately. The next most important change is the auto mode fix: the AskUserQuestion suppression bug — which had previously caused necessary confirmations to be silently dropped during automated runs — has been resolved. The behavior now ensures that AskUserQuestion is not suppressed when the user or a skill explicitly depends on it, making this a good time to revisit workflow designs that rely on auto mode.
On the security and stability side, the "command line is invalid" regression affecting winget / Microsoft Store installations on Windows PowerShell since v2.1.124 has been fixed. This had been a blocker for team-wide deployments on Windows, and distribution via winget upgrade is now usable again. The bug where pages 2 and beyond were dropped in MCP's resources/list, resources/templates/list, and prompts/list has also been fixed — entries that were previously missed when connecting to large MCP servers with 20+ resources or prompts can now be fully retrieved.
For enterprise users, the issue where forceLoginOrgUUID / forceLoginMethod policies were not enforced for third-party provider / API key sessions has been resolved. Access controls configured via Managed Settings, which previously passed through unchecked for Bedrock, Vertex, or direct API key sessions, are now properly enforced, making internal SSO governance effective. For multi-agent deployments, forwarding of CLAUDE_CODE_SUBAGENT_MODEL to child processes has been fixed, so designs that specify a different model for subagents now work as intended.
On the UX side, the following issues have all been resolved: right-click / middle-click paste not inserting content in GNOME Terminal, a full-screen strobe effect during background sessions in Windows Terminal, /background refusing to start with only a skill or custom slash command as input, previously approved "don't ask again" permissions being re-prompted in background sessions, and the /theme color editor and "New custom theme" dialog not closing with the Escape key. Diff rendering for large file edits has been sped up, the auto-updater now displays the current version in the status line on update failure, and native version checking plus automatic retry on temporary network failures have been improved. A guard has also been added to prevent background job cleanup in Windows worktrees from traversing NTFS junctions into the main repo.
| Change Category | Details | Affected Area |
|---|---|---|
| Command rename | /simplify → /code-review, with effort level argument support |
CI scripts, runbooks, team documentation |
| Auto mode fix | AskUserQuestion suppression lifted when user/skill depends on it |
All automated workflows |
| MCP pagination | Retrieval of pages 2+ in resources/list / templates/list / prompts/list |
Large MCP server integrations |
| Access control | forceLoginOrgUUID / forceLoginMethod applied to third-party provider / API key sessions |
Enterprise SSO governance |
| Windows PowerShell | Fixed "command line is invalid" for winget / Microsoft Store installations | Windows team deployments |
| GNOME / Windows Terminal | Fixed right-click / middle-click paste and background strobe | Linux / Windows professional use |
[2] Code w/ Claude London Day 1–Day 2 Event Report — On-Site Coverage by Fortune and MIT Tech Review
We're reporting on Code w/ Claude London (May 20–21, 2026), an Anthropic-hosted developer conference, which was covered on-site by Fortune and MIT Technology Review on May 21, 2026 local time (primary sources: https://fortune.com/2026/05/21/claude-code-london-anthropic-ai-software-engineering/ , https://www.technologyreview.com/2026/05/21/1137735/anthropics-code-with-claude-showed-off-codings-future-whether-you-like-it-or-not/ , https://www.anthropic.com/code-with-claude ).
Impact for engineers: ★★
According to the coverage, the Day 1 keynote provided detailed explanations of the four items from the May 19 release notes (MCP Tunnels research preview, Managed Agents self-hosted Sandbox, live MCP server configuration updates, and automatic spilling for outputs exceeding 100K tokens) — no additional APIs or new product announcements were confirmed. Customer case study sessions were also held featuring Asana, Cursor, GitHub, Replit, and Vercel, sharing each company's use cases for Claude Code adoption. Claude for Small Business (released May 13, 2026) was also re-introduced at the venue.
The industry context commentary from external media (Fortune and MIT Technology Review) is useful as supplementary material for enterprise-facing presentations and as a source for articles on the "mainstreaming" of AI coding. MIT Technology Review in particular titled its piece "whether you like it or not," framing AI coding as an inevitable and entrenched trend, making it a prime citation candidate for internal adoption proposals. Registration for the Tokyo venue (June 5–6, 2026) is still ongoing, and engineers in Japan are advised to confirm their registration promptly.
[What to Watch Next] Immediate Adoption of v2.1.146 and /code-review Name Replacement
In the editorial team's assessment, the top-priority tasks to tackle immediately are updating to Claude Code v2.1.146 and replacing /simplify with /code-review in CI scripts and runbooks. The former includes fixes directly tied to operational stability — MCP pagination, forceLoginOrgUUID, and the auto mode AskUserQuestion bug — and prompt updates to both CI and developer environments are recommended. The latter is critical because leaving the name unreplaced will cause automated review jobs in CI to silently break, so the safe approach is to do a grep-and-replace across the codebase on the day of or the day after the release notes are published. We will continue tracking any Day 3+ updates from Code w/ Claude London and on-site reports from the Tokyo venue (June 5–6) in this news section.
[Sources] Primary Source Links Referenced in This Article
- Claude Code Releases (v2.1.146): https://github.com/anthropics/claude-code/releases/tag/v2.1.146
- Claude Code Changelog: https://code.claude.com/docs/en/changelog
- Code w/ Claude London Report (Fortune): https://fortune.com/2026/05/21/claude-code-london-anthropic-ai-software-engineering/
- Code w/ Claude London Report (MIT Technology Review): https://www.technologyreview.com/2026/05/21/1137735/anthropics-code-with-claude-showed-off-codings-future-whether-you-like-it-or-not/
- Code w/ Claude Official: https://www.anthropic.com/code-with-claude
- Anthropic News (top): https://www.anthropic.com/news
- Platform Docs Release Notes: https://platform.claude.com/docs/en/release-notes/overview
