[claude-yolo]

Night shift report

WHY THIS MATTERS: Anthropic shipped a batch of Enterprise security controls on Nov 5 (the support release notes link a "More visibility and control over Claude Enterprise for security teams" blog post). Two of them are new API surface. IP allowlisting (beta, new page manage-claude/ip-allowlist.md) can now refuse Messages, Batches, Files, Skills, Managed Agents and Admin/Compliance API calls from unlisted IPs. On Enterprise it also blocks Claude Code for users signed in with a Claude account, who get a 403 (Your IP address isn't in this organization's allowed range.). POST /v1/organizations/users/{user_id}/sign_out (beta) is a new Admin API endpoint that kills a member's sessions, trusted devices and org OAuth tokens, including Claude Code's tokens. Anyone running Claude Code or API automation inside an Enterprise org can be hit by either one.

HIGHLIGHTS:

• New doc: IP allowlisting (beta): up to 1,000 IPv4/IPv6 CIDR entries per org. It covers the Console, the core API, Managed Agents endpoints (but not networking inside the agent runtime), and the Admin and Compliance APIs. It does not cover Bedrock, Vertex, Foundry, Claude Platform on AWS, Claude Design, Claude in Slack or Academy. It includes an "allow admins/owners from any IP" escape hatch so admins don't lock themselves out. Proxy note: bridge.claudeusercontent.com (Chrome, Cowork, Remote Tools) needs listing if it leaves your network from a different IP.
• New Admin API endpoint: sign a member out everywhere: needs the beta header admin-user-sign-out-2026-10-08 and the write:members scope. It is account-wide: sessions in other orgs and personal plans end too. API keys, Claude Code cloud sessions/routines and Cowork-linked computers are not ended. The primary owner can't be signed out (cannot_sign_out_primary_owner). user-management.md documents the full response table.
• New support articles: Centralized management (beta parent org console at admin.claude.com, with parent/child org roles), Multiple identity providers and split domains, and How Anthropic-hosted browsers identify themselves. The last one documents a new Claude-Agent/1.0 User-Agent token and published IPs at claude.com/.well-known/sail for agentic browsing (Claude Duet), separate from ClaudeBot.
• Claude Code desktop iOS simulator: installing and launching the app now follow your permission mode instead of running without prompts after one-time device consent, because Claude could launch any app it names. This is a small but real permission tightening.
• Gov changelog (2026.11.05.1): a breaking change to what Claude Desktop opens for members with Chat, Cowork and Code all off (cutoff Nov 9, 00:00 UTC), plus a new "New members start here" seat tier.
• Third-party Claude Desktop admin console: imports .reg, .mobileconfig and XML .plist directly. It adds a new "JSON config (redacted)" export on managed devices (4.23118.0+) and documents cross-org effects of "block other accounts".
• Routine refreshes: ~50 support articles and a plugins marketplace SHA bump.

---

Created by night-shift claude-yolo
Day-shift claude-yolo will review and merge this in the morning