Claude Code Weekly Updates (2.1.51 - 2.1.63)
Here are the Claude Code updates from February 23 - March 1, 2026 that caught my attention.
TL;DR
- version: 2.1.51 -> 2.1.63
- changes: 86+
- Important topics
- Remote Control (
claude remote-control) - /simplify & /batch commands
- Auto-memory
- Remote Control (
v2.1.51 (11 changes)
Remote Control
Added
claude remote-controlsubcommand for external builds, enabling local environment serving for all users.
New in Claude Code: Remote Control. Kick off a task in your terminal and pick it up from your phone while you take a walk or join a meeting. Claude keeps running on your machine, and you can control the session from the Claude app or claude.ai/code
A feature that lets you operate your local Claude Code session from a mobile device or Web UI. It works similarly to local hosting.
You can start it with claude remote-control or /remote-control.
By setting “Enable Remote Control for all sessions: true” in /config, all local sessions become accessible from the mobile app or browser.
Personally, I think this feature was developed with OpenClaw in mind. As long as your machine stays powered on and connected to the network, you can access your Claude Code session from anywhere.
v2.1.52 & v2.1.53 (9 changes)
v2.1.52 was only a VS Code crash fix, and v2.1.53 was primarily a bug fix release. Nothing particularly noteworthy in my opinion.
v2.1.55, v2.1.56 & v2.1.58 (3 changes)
v2.1.55, 56, and 58 are also mainly bug fixes, so nothing particularly noteworthy in my opinion.
Remote Control has now been rolled out to Pro plan users as well.
v2.1.59 (8 changes)
Auto-memory
Claude automatically saves useful context to auto-memory. Manage with /memory
Claude now automatically saves useful context to memory.
You can toggle it on/off and review saved content with the /memory command. Per-subagent memory from previous sessions is also accessible through /memory.
An auto-memory folder (~/.claude/projects/<proj-dir>/memory/) was added, enabling persistent per-project memory. I’ve already seen cases where it can cause issues, but I’m hopeful it will become a great feature over time.
You can disable Auto-memory with CLAUDE_CODE_DISABLE_AUTO_MEMORY=1.
/copy Command Enhancement
Added
/copycommand to show an interactive picker when code blocks are present, allowing selection of individual code blocks or the full response.
Previously you could only copy the full response, but now you can select and copy individual code blocks.
The code block labels also make it easier to identify which language each code block is in.
You can revert to the previous full-response-only behavior by setting “Always copy full response (skip /copy picker): true” in /config.
v2.1.61 & v2.1.62 (2 changes)
Cache Regression Fix
Fixed prompt suggestion cache regression that reduced cache hit rates
v2.1.62 fixed a prompt suggestion cache regression. This issue caused usage limits to be consumed faster than usual, affecting many users. Anthropic also performed rate limit resets.
Claude Codeのレートリミットがリセットされました。 先日入ったprompt cachingの変更の影響で、usage limitの消費が早くなる問題が原因とのこと。2.1.62では修正されているようです。
We've reset rate limits for all Claude Code users. Yesterday we rolled out a bug with prompt caching that caused usage limits to be consumed faster than normal. This is hotfixed in 2.1.62. Make sure you upgrade to the latest and hope you enjoy using Claude Code this weekend!
v2.1.63 (26 changes)
/simplify & /batch Commands
Added
/simplifyand/batchbundled slash commands
Two skills that the Claude Code development team uses internally have been publicly released: /simplify and /batch.
- /simplify: A skill where 3 agents review changed code in parallel and fix any issues on the spot.
- /batch: A skill that splits large, parallelizable changes across multiple worker agents, each running simultaneously and creating PRs automatically.
Let’s take a closer look at each.
/simplify
Overview: A skill that automatically reviews changed code from 3 perspectives and fixes issues on the spot.
3-Phase Architecture
Phase 1: Identify Changes
-
Gets changed content via git diff (if no changes, targets recently edited files)
▼
Phase 2: 3-Agent Parallel Review
-
Agent 1: Code Reuse Review “Are you writing new code when existing utilities could handle it?” “Is there duplicate logic with functions elsewhere?”
-
Agent 2: Code Quality Review “Is there redundant state management?” “Are there copy-pasted blocks with subtle differences?” “Are there hardcoded strings when constants exist?”
-
Agent 3: Efficiency Review “Are there unnecessary computations or API calls?” “Is something running sequentially when it could be parallelized?” “Are you fetching all records when only one is needed?”
▼
Phase 3: Fix Consolidates all agent results and directly fixes issues (skips false positives)
/batch
Overview: A skill that splits large, parallelizable changes across multiple worker agents, each running simultaneously and creating PRs automatically.
3-Phase Architecture:
Phase 1: Research and Planning (Plan Mode) 1-1. Scope Understanding: Investigate codebase with Explore agents 1-2. Decompose into Independent Units: Break into 5-30 self-contained work units 1-3. Determine E2E Tests: Finalize verification methods (asks user if none found) 1-4. Create Plan File: Unit list + worker instruction template 1-5. Request approval via ExitPlanMode
▼
Phase 2: Launch Workers After approval, spawn one agent per unit
-
All workers use isolation: “worktree” for independent copies
-
All workers use run_in_background: true for background execution
-
All workers launched simultaneously in a single message, each performing:
- Implementation
- Self-review with /simplify
- Unit test execution
- E2E verification
- Commit, push, and PR creation
▼
Phase 3: Progress Tracking Real-time status table updates PR links recorded as completion notifications arrive Final summary displayed after all workers complete
Use Cases
- Library major version upgrades: Rewriting imports across all files
- Naming convention changes: snake_case to camelCase, etc.
- Design pattern migrations: class to hooks, callback to async/await
- Adding multi-language support: Adding i18n keys to each screen
- Bulk test additions: Creating test files per module
HTTP hooks
Added HTTP hooks, which can POST JSON to a URL and receive JSON instead of running a shell command
A new http type has been added to hooks, which previously only supported command, prompt, and agent types. HTTP hooks add a new transport layer to the existing hooks system.
HTTP hook fields are url, headers, and allowedEnvVars. Unlike command hooks where exit 2 explicitly blocks execution, HTTP hooks use a “fail-open” design since network failures can’t be distinguished from intentional blocks. To block an action, you must return a 2xx response with a block instruction. For enterprise deployments, admins can restrict allowed URLs via allowedHttpHookUrls.
Other Claude Code Topics
Claude Code Mobile App Supports Plan Mode
By updating the mobile app, Plan mode and AskUserQuestion are now available.
Claude Cowork Scheduled Tasks
Claude Cowork がスケジュールタスクをサポート マシンの電源をオンにしておく必要はありますが、定期実行などが可能になります。
New in Cowork: scheduled tasks. Claude can now complete recurring tasks at specific times automatically: a morning brief, weekly spreadsheet updates, Friday team presentations.
Claude Cowork now supports scheduled tasks. Your machine needs to stay powered on, but periodic execution is now possible.
This also seems to be developed with OpenClaw in mind.
Claude for Open Source Program
Claude for Open Source プログラムが開始 ・OSSのメンテナーやコアコントリビューターに、Claude Max 20xを6ヶ月間無料でプレゼント ・メイン/コアメンテナーの場合: Public GitHubリポジトリで5,000 stars以上、または NPMパッケージで月間1M（100万）ダウンロード以上
Excited to announce Claude for Open Source ❤️ We're giving 6 months of free Claude Max 20x to open source maintainers and core contributors. If you maintain a popular project or contribute across open source, please apply! claude.com/contact-sales/…
A program offering Claude Max 20x free for 6 months to OSS maintainers and core contributors has launched. For main/core maintainers: 5,000+ stars on public GitHub repos, or 1M+ monthly NPM downloads. Others can apply if they maintain ecosystem-critical projects.
Anthropic vs Pentagon
Anthropic CEOのDarioとペンタゴンとの間で起きている、軍事目的でのAI利用を巡る対立について Anthropicは、軍事利用の98〜99%については同意しているが、民主主義的な価値観を守るために、2つのユースケースについては制限を設けている ・国内の大量監視（Domestic Mass Surveillance）:
A clash between Anthropic CEO Dario and the Pentagon over military use of AI is unfolding. Please refer to the post above for details, but personally, I think this could become a turning point in history.
As a result of these events, Claude reached #1 on the US App Store.
Closing
Thank you for reading!