What's new in Claude

Added claude mcp login/logout CLI commands for MCP server auth without interactive menus; !-prefixed bash commands now auto-trigger Claude responses. Also: f key to filter by status in /workflows, iTerm2 teammate mode support, and fixes for post-sleep streaming errors.

Updated stream-stall hint messaging and adjusted retry timing from 10s to 20s of silence before triggering, reducing false-positive retries on slow network connections.

Improved auto mode safety: destructive git commands and infrastructure-destroy operations are now blocked unless explicitly requested, preventing accidental data loss during autonomous runs.

/cd command: Move the current session to a new working directory mid-conversation without rebuilding the prompt cache — useful when switching between repos in a long session.

Sub-agents can spawn sub-agents: Background chains are now supported up to five levels deep, enabling more complex multi-agent orchestration patterns out of the box.

--safe-mode flag: Starts Claude Code with all customizations disabled (hooks, plugins, settings overrides) for clean troubleshooting when something in your config is misbehaving.

fallbackModel config: Configure up to three fallback models tried in order when your primary model is unavailable or rate-limited, keeping long runs from failing silently.

Model retirements: Claude Sonnet 4 (claude-sonnet-4-20250514) and Claude Opus 4 (claude-opus-4-20250514) are retired — all requests now return an error. Upgrade to Sonnet 4.6 or Opus 4.8 respectively.

Tool updates: Code execution tool now supports code_execution_20260521 with explicit 90s per-cell time limit. Web search and web fetch tools add a response_inclusion parameter to drop consumed result blocks from API responses for agentic workflows. No beta header required for either.

Claude Fable 5 launched: Most capable widely-released model — 1M token context, 128k max output, always-on adaptive thinking, and a new tokenizer (~30% more tokens vs. pre-Opus 4.7). Requires 30-day data retention; not available under ZDR. Claude Mythos 5 also launched for Project Glasswing participants.

Claude Code Artifacts is live for Team and Enterprise plans — turns any Claude Code session into a live, shareable HTML page (dashboards, PR walkthroughs, release checklists). Also shipped: flexible prompt-time config commands and new macOS sandbox + notification controls.

Anthropic Artifacts update brings live, shared dashboards and interactive workspaces to enterprises. Plug in live code, multiple data sources, and surface work on an interactive URL directly from your Claude Code session.

Boris (Claude Code creator) on his setup: "Claude Code works great out of the box — I personally don't customize it much. There is no one correct way to use Claude Code." Also noted he hasn't written a line of code himself in 2026.

Thariq shipped "Todos into Tasks" in Claude Code — upgrading the task system. Also posted about dynamic workflows: Claude now writes its own harness on the fly, custom-built for the task at hand, enabling dozens to hundreds of subagents from a single script.