Claude Code Week 16 (April 13–17, 2026): Opus 4.7, xhigh Effort, Routines, /ultrareview, /usage, Native Binaries

Anthropic's Week 16 update (April 13–17, 2026, v2.1.105 → v2.1.113) is the biggest Claude Code shipment of the month. Claude Opus 4.7 lands as the new default on Max and Team Premium with a new xhigh effort level that's the recommended setting for most coding work, Routines turn cloud agents into something you schedule or trigger from GitHub events, /ultrareview runs parallel multi-agent code review in the cloud, /usage shows what's actually driving your limits, and the CLI quietly moves to native per-platform binaries.

TL;DR

• Claude Opus 4.7 is the new default on Max and Team Premium; available everywhere else from /model.
• xhigh effort level sits between high and max; applied by default the first time you switch to 4.7. /effort opens an interactive slider when called without arguments.
• Routines (web): templated cloud agents triggered by schedule, GitHub events, or a tokened /fire API endpoint.
• /ultrareview (v2.1.111): parallel multi-agent code review in the cloud, with an adversarial critique pass.
• /usage: a breakdown of parallel sessions, subagents, cache misses, and long context driving your limits, with d/w for day/week views.
• Native binaries (v2.1.113): claude is now a per-platform native binary; npm pulls it in via an optional dependency.

Quick Answer

Update Claude Code (claude update) to pick up v2.1.113 and the native binary. Run /model opus to confirm you're on Opus 4.7, then /effort to land on xhigh. Run /usage to see what's eating your limits, and /ultrareview from a branch you're about to merge for a parallel cloud-side review.

What Shipped in Week 16

Claude Opus 4.7 and the xhigh Effort Level

Opus 4.7 is now the default on Max and Team Premium and is selectable everywhere else from /model. The notable addition for power users is the new xhigh effort level: a slot between high and max that Anthropic recommends as the best setting for most coding and agentic tasks. The first time you switch to 4.7, xhigh is applied as the default, so you don't have to discover it manually.

/effort without arguments now opens an interactive arrow-key slider, which removes the "what was the level called again?" tax of bouncing between effort presets.

> /model opus
> /effort xhigh

A separate Week 17 fix (covered in our Week 17 update) computes Opus 4.7 sessions against the model's native 1M-token context window, fixing inflated /context percentages and premature autocompaction.

If you're weighing 4.7 against earlier models, our Claude Opus 4.6 vs Opus 4.5 comparison is the closest reference for how Anthropic sequences these releases.

Routines: Templated Cloud Agents With Real Triggers

Routines are templated cloud agents you define once on Claude Code on the web with a prompt, the repos they can touch, and the connectors they need — then a schedule, a GitHub event (PR opened, release published, with optional filters), or your own webhook triggers them without your machine running. The trigger picker now covers GitHub events with optional filters, and every routine gets a tokened /fire endpoint so external systems can kick one off.

You can scaffold a routine from the CLI:

> /schedule daily PR review at 9am

This is the natural next step after Week 13's /autofix-pr and Week 15's /ultraplan: instead of running cloud agents one-shot from your terminal, you teach a recurring or event-driven version once and wire it into the rest of your tooling.

/ultrareview: Parallel Multi-Agent Code Review

/ultrareview runs comprehensive code review in the cloud. It fans your branch out across parallel reviewers on Claude Code on the web, runs an adversarial critique pass over each finding to drop noise, and returns a verified findings report while your terminal stays free. The launch dialog now shows a diffstat so you know exactly what's going up before you confirm.

Review the branch you're on:

> /ultrareview

Or point it at a PR number:

> /ultrareview 1234

/ultrareview opened to public research preview in Week 17. It's the cloud-side counterpart to running code review tools locally.

/usage: Where Your Limits Actually Go

/usage now shows a breakdown of what's driving your limits: parallel sessions, subagents, cache misses, and long context, each with a percentage of the last 24 hours and a tip to optimize it. Press d for the day view or w to switch to a week view.

> /usage

This is more useful than a raw token counter — it tells you whether the bill is coming from running many sessions in parallel, from subagents, from cold cache, or from carrying a long context that should have been compacted. In Week 17, /cost and /stats were merged into /usage, with the old names kept as typing shortcuts that open the relevant tab.

Native Binaries Replace Bundled JavaScript

The claude CLI now spawns a native per-platform binary instead of bundled JavaScript, so the installed claude command no longer invokes Node. The npm package pulls the right binary in through an optional dependency such as @anthropic-ai/claude-code-darwin-arm64, so your install command doesn't change. The standalone installer already shipped this binary; npm now matches it.

claude update
claude --version

For most users this is invisible, but startup time improves and there's no more Node version drift between machines.

Other Notable Wins in Week 16

A handful of smaller changes worth knowing about:

• Auto mode is now available for Max subscribers on Opus 4.7, and the --enable-auto-mode flag is no longer required.
• Session recap ships a one-line summary of what happened while you were away; run /recap on demand or turn it off from /config. (Expanded in Week 17.)
• New /tui command and tui setting switch between classic and flicker-free rendering mid-conversation; focus view moved from Ctrl+O to its own /focus command.
• Push notification tool: with Remote Control connected and "Push when Claude decides" enabled, Claude can ping your phone when it needs you.
• Plugins can ship background watchers via a top-level monitors manifest key that auto-arms at session start or on skill invoke.
• An "Auto (match terminal)" option in /theme follows your terminal's dark/light mode.
• /fewer-permission-prompts scans your transcripts for common read-only Bash and MCP calls and proposes an allowlist for .claude/settings.json.
• Claude can now discover and run built-in commands like /init, /review, and /security-review via the Skill tool.
• PreCompact hooks can block compaction by exiting with code 2 or returning {"decision":"block"}.
• ENABLE_PROMPT_CACHING_1H opts API key, Bedrock, Vertex, and Foundry users into a 1-hour prompt cache TTL.
• sandbox.network.deniedDomains carves specific domains out of a broader allowedDomains wildcard.
• /undo is now an alias for /rewind, and /proactive is an alias for /loop.
• Hardened Bash permissions: deny rules now match through env/sudo/watch wrappers, and Bash(find:*) allow rules no longer auto-approve -exec or -delete.

How This Compares to Other Agents

For deeper context, see our Claude Code vs Aider guide, the cursor-vs-claude-code">Cursor vs Claude Code comparison, and the Claude Code tool page.

Sources

• Claude Code "What's New" — Week 16 (official): code.claude.com/docs/en/whats-new/2026-w16
• Claude Code "What's New" index: code.claude.com/docs/en/whats-new
• Claude Code changelog v2.1.105 → v2.1.113: code.claude.com/docs/en/changelog
• Ultrareview guide: code.claude.com/docs/en/ultrareview
• Routines guide: code.claude.com/docs/en/routines
• Model config — effort levels: code.claude.com/docs/en/model-config#adjust-effort-level

For more context on recent Claude Code shipments, see our Week 15 update and Week 17 update. For a broader read on agent capabilities, the AI coding agents explainer is a good starting point.