What's new in Claude

Stacked slash-skill invocations now load all leading skills (up to 5), SSL certificate errors fail immediately with actionable guidance, and streaming responses with mid-stream errors now preserve partial output with an incomplete-response notice.

Subagents now run in the background by default with Claude continuing to work while they run, Claude in Chrome is now generally available, and background agents automatically commit, push, and open draft PRs when finishing code work.

claude mcp login: authenticate a configured MCP server directly from your shell instead of navigating the interactive /mcp menu, and clear stored credentials later with claude mcp logout.

Shell mode responds to command output: running ! npm test now gets an explanation from Claude without requiring a second prompt.

/rewind can resume from before /clear: you can now jump back to a conversation state that existed before a /clear was run, recovering context you thought was gone.

Background subagent permission prompts: subagents running in the background now surface permission prompts in the main session instead of auto-denying, so long-running tasks no longer silently fail on permissions.

Anthropic restored access to Claude Fable 5 and Claude Mythos 5 following a brief suspension — see the Anthropic statement for full details.

Claude Sonnet 5 (claude-sonnet-5) launched at introductory pricing of $2/$10 per MTok through August 31, 2026 (standard $3/$15 after). Supports 1M token context, 128k max output, and adaptive thinking on by default. Note: the new tokenizer produces ~30% more tokens for the same text.

Managed Agents event deltas: session event streams now support streaming deltas so you can preview agent message text as it generates, before the complete event arrives — opt in with the event_deltas[] query parameter.

Rate limits raised: Claude Sonnet and Haiku rate limits now match Claude Opus across all tiers, and usage tiers consolidated from many down to three — Start, Build, and Scale. No action required; no organization receives lower limits than before.

Sonnet 5 is a clear upgrade from 4.6, and the claude-api skill makes migration easier by tuning prompts for Sonnet 5, recommending effort levels, and configuring advisor mode.

Doubled Claude Cowork usage limits for a month, applying to 5-hour rate limits. "If you've been saving up a big messy project, now's the time."

Shared a deep-dive on dynamic workflows in Claude Code — Claude writes its own harness on the fly, custom-built for the task at hand, enabling complex multi-agent orchestration without manual wiring.