Claude Code Weekly Updates (2.1.77 - 2.1.81)

Posted on March 22, 2026

title

Here are the Claude Code updates from March 16 - 22, 2026 that caught my attention.

TL;DR

version: 2.1.77 -> 2.1.81
changes: ~128
Important topics
- Opus 4.6 output token limit raised to 64k (max 128k)
- Skill .md command output embedding with !command
- StopFailure hook event added
- Claude Code channels (Telegram/Discord) research preview
- --bare flag for scripting
- Statusline rate_limits field added

v2.1.77 (41 changes)

v2.1.77

A version focused on the major Opus 4.6 output token limit expansion and several tmux/iTerm2 fixes.

Opus 4.6 Output Token Limit Increase

Increased default max output tokens for Claude Opus 4.6 to 64k tokens, and raised the upper bound for Opus 4.6 and Sonnet 4.6 models to 128k tokens

The default max output tokens for Claude Opus 4.6 has been raised to 64k, and the upper bound for both Opus 4.6 and Sonnet 4.6 models has been increased to 128k tokens.

Previously, Claude Code internally capped the default at 32k with an upper bound of 64k. Since Opus 4.6 supports up to 128k at the API level, the CLI settings were the bottleneck.

For reference, here are the API-level max output tokens across model generations:

Model	Max Output
Claude Opus 4 / 4.1	32k
Claude Opus 4.5	64k
Claude Opus 4.6	128k

Models overview

Claude is a family of state-of-the-art large language models developed by Anthropic. This guide introduces the available models and compares their performance.

platform.claude.com

`/fork` Renamed to `/branch`

Renamed /fork to /branch

The /fork command has been renamed to /branch, aligning more closely with the git branch concept.

v2.1.78 (26 changes)

v2.1.78

`StopFailure` Hook Event

Added StopFailure hook event that fires when a turn ends due to API errors (rate limits, authentication failures, etc.)

A new hook event that fires instead of Stop when a turn ends due to an API error. Normal completions fire Stop, while API errors fire StopFailure.

You can use matchers to filter by error type and handle specific errors:

matcher	Description
`rate_limit`	Rate limit exceeded
`authentication_failed`	Authentication error
`billing_error`	Billing/account issue
`server_error`	Server-side error
`max_output_tokens`	Output token limit exceeded

The hook payload includes error (error type), error_details (details), and last_assistant_message (the displayed error message). Unlike Stop, StopFailure has no decision control. Output and exit codes are ignored, so it is intended for side effects like notifications and logging.

Sample:

{
  "session_id": "abc123",
  "transcript_path": "/Users/.../.claude/projects/.../00893aaf-19fa-41d2-8238-13269b9b3ca0.jsonl",
  "cwd": "/Users/...",
  "hook_event_name": "StopFailure",
  "error": "rate_limit",
  "error_details": "429 Too Many Requests",
  "last_assistant_message": "API Error: Rate limit reached"
}

Hooks reference - Claude Code Docs

Reference for Claude Code hook events, configuration schema, JSON input/output formats, exit codes, async hooks, HTTP hooks, prompt hooks, and MCP tool hooks.

code.claude.com

Frontmatter effort/maxTurns/disallowedTools

Added frontmatter support for effort, maxTurns, and disallowedTools in skills and slash commands

Skills and slash commands now support effort, maxTurns, and disallowedTools in their frontmatter.

Field	Description
`effort`	Sets the effort level for skill execution: `low` / `medium` / `high` / `max` (`max` is Opus 4.6 only). Overrides the session effort setting
`maxTurns`	Maximum number of agent turns. Prevents skills from running indefinitely
`disallowedTools`	A denylist for tools. The inverse of the existing `allowed-tools` (allowlist), enabling a “allow everything except these specific tools” pattern

effort is particularly practical. You can set low on lightweight skills to reduce token consumption, or max on complex analysis skills to enable maximum reasoning. disallowedTools was already supported for subagents and has now been extended to skills as well.

Extend Claude with skills - Claude Code Docs

Create, manage, and share skills to extend Claude's capabilities in Claude Code. Includes custom commands and bundled skills.

code.claude.com

v2.1.79 (18 changes)

v2.1.79

Show Turn Duration Toggle

Added “Show turn duration” toggle to /config menu

A toggle has been added to the /config menu to display how long each turn takes.

turn

`/remote-control` for VSCode

Added /remote-control for VSCode

Remote control is now available from VSCode. You can send task instructions from the web or mobile apps to your VSCode environment.

VSCode feature implementations tend to lag slightly behind the CLI, but the gap has been narrowing noticeably.

AI-Generated Session Titles

Added AI-generated session titles

Session titles are now automatically generated by AI based on the conversation context and plan content.

After about three turns of conversation, a session name is automatically assigned. Very convenient.

session-name

v2.1.80 (17 changes)

v2.1.80

Statusline `rate_limits` Field

Added rate_limits field to statusline scripts with 5-hour and 7-day window rate limit status, including used_percentage and resets_at

A rate limit status field has been added for statusline scripts. You can display usage percentage (used_percentage) and reset time (resets_at) for both 5-hour and 7-day windows.

The following article provides a clear guide on how to configure this:

Claude Codeの使用率がステータスラインに表示できるようになったので表示用のスクリプトを作った話

Claude Code v2.1.80で追加されたrate_limitsフィールドを使って、5時間/7日間の使用量をステータスラインにかわいく表示する4つのパターンを紹介します

nyosegawa.com

`--channels` Research Preview

Added --channels research preview for MCP server message pushing to mobile devices

Claude Code channels has been released as a research preview. This enables using Claude Code from Telegram and Discord. Clearly inspired by OpenClaw.

I wrote about how to set up Discord integration in the following post:

Oikon

@oikon48

· Follow

【Claude Code を Discord から使う方法】 Claude CodeをTelegram, Discordから実行できるよういん (2.1.80~) 。試してみたのでやり方をまとめる。 1. Discord Developer Portal でアプリ作成: ( discord.com/developers/app… ) 2. Bot を作成してToken取得: (Bot -> Reset Token -> Copy) 3. Message

11:59 PM · Mar 19, 2026

572

Read 11 replies

I tried both Telegram and Discord integration, and Telegram is easier to set up.

Both can be installed from Anthropic’s official plugin marketplace.

`effort` Frontmatter for Skills

Added effort frontmatter support for skills and slash commands

The effort level can now be specified in skill and slash command frontmatter.

v2.1.81 (26 changes)

v2.1.81

`--bare` Flag

Added --bare flag for scripted -p calls that disables hooks, LSP, plugin sync, skill directory scanning. API key must come from ANTHROPIC_API_KEY or apiKeyHelper in --settings (OAuth and keychain are not available). Auto-memory is off.

The --bare flag has been added for scripted -p mode. Here’s how it compares to regular -p mode:

Feature	`-p` (regular)	`-p --bare`
hooks	Enabled	Disabled
LSP	Enabled	Disabled
Plugin sync	Enabled	Disabled
Skill directory scanning	Enabled	Disabled
OAuth / Keychain auth	Enabled	Disabled
Auto-memory	Enabled	Disabled

API keys can only be provided via the ANTHROPIC_API_KEY environment variable or apiKeyHelper in --settings. By skipping initialization of hooks and other subsystems, this is useful for speeding up startup in CI/CD pipelines and simple scripts.

`--channels` Permission Relay

Added --channels permission relay to forward tool approvals to mobile devices

Tool approval forwarding to channel servers has been added. You can now approve Claude Code tool usage from mobile devices.

Other Claude Code Topics

Command Output Embedding in Skill .md

Skill command embedding

It was discovered that writing !`command` in a Claude Code Skill .md file embeds the command’s execution output directly when the skill is expanded. This feature is not documented in the CHANGELOG.

This enables embedding dynamic context information (e.g., current git status, environment info) into skills. The allowed-tools frontmatter field must include the necessary tools.

Anthropic Skills Article

The Anthropic team published “Lessons from Building Claude Code: How We Use Skills,” categorizing skills and sharing tips for accelerating development.

The article classifies skills into categories such as library/API references, workflow automation, and code review/testing. It also introduces using PreToolUse hooks to measure team skill usage rates.

Thariq

@trq212

· Follow

Using Skills well is a skill issue. I didn't quite realize how much until I wrote this, the best can completely transform how your team works.

Thariq

@trq212

x.com/i/article/2033…

5:29 PM · Mar 17, 2026

3.8K

Read 90 replies

everything-claude-code Surpasses Claude Code’s Star Count

Oikon

@oikon48

· Follow

everything-claude-codeが、本家のClaude Codeのスター数を抜いたんか

cogsec

@affaanmustafa

May I have your attention, please? Will the real Claude Code (repo) please stand up? I repeat Will the real Claude Code (repo) please stand up?

12:33 AM · Mar 17, 2026

256

Read 1 reply

The everything-claude-code repository’s star count has surpassed that of the official Claude Code repository. I recently wrote an article about Everything Claude Code, so feel free to check it out:

What can be learned from Everything Claude Code | oikon48

Lessons from Everything Claude Code, the repository from the Anthropic hackathon winner.

oikon48.dev

Claude Dispatch

Oikon

@oikon48

· Follow

Claude のモバイルアプリに ”Dispatch” が追加されていますね👀 Claude Desktopと接続してタスクをCoworkに投げることができます。Claude Codeでいうremote controlみたいなやつ。

4:01 AM · Mar 18, 2026

Read 2 replies

“Dispatch” has been added to Claude’s mobile app. It connects to Claude Desktop and lets you send tasks to Cowork. It’s similar to Claude Code’s remote control feature.

Notably, Claude Dispatch includes a feature that prevents your machine from sleeping during execution, making dedicated desktop machines or always-on environments even more important.

Claude Cowork Projects

Projects are now available in Claude Cowork.

Oikon

@oikon48

· Follow

Claude Cowork にProjects機能が使用可能に👀

Claude

@claudeai

Projects are now available in Cowork. Keep your tasks and context in one place, focused on one area of work. Files and instructions stay on your computer. Import existing projects in one click, or start fresh.

5:16 PM · Mar 20, 2026

Read 1 reply

Thank you for reading!

TL;DR

v2.1.77 (41 changes)

Opus 4.6 Output Token Limit Increase

/fork Renamed to /branch

v2.1.78 (26 changes)

StopFailure Hook Event

Frontmatter effort/maxTurns/disallowedTools

v2.1.79 (18 changes)

Show Turn Duration Toggle

/remote-control for VSCode

AI-Generated Session Titles

v2.1.80 (17 changes)

Statusline rate_limits Field

--channels Research Preview

effort Frontmatter for Skills

v2.1.81 (26 changes)

--bare Flag

--channels Permission Relay

Other Claude Code Topics

Command Output Embedding in Skill .md

Anthropic Skills Article

everything-claude-code Surpasses Claude Code’s Star Count

Claude Dispatch

Claude Cowork Projects

`/fork` Renamed to `/branch`

`StopFailure` Hook Event

`/remote-control` for VSCode

Statusline `rate_limits` Field

`--channels` Research Preview

`effort` Frontmatter for Skills

`--bare` Flag

`--channels` Permission Relay