Community posts

Anthropic opened Claude Code to mods. These are small TypeScript hooks that can rewrite prompts, swap UI, or replace built-ins. They're not sandboxed. They have the same machine access as Claude Code. Install only from sources you trust, or have Claude write the mod for you. The right bet is to turn the coding agent into a platform, not a black box. https://claude.com/blog/claude-code-mods

JUST IN 🚨 CLAUDE CODE MODS ARE LIVE Anthropic — TypeScript hooks that rewrite prompts, tool calls, and UI. Swap built-ins like /diff. CLI + desktop. Ships inside plugins. 🔗 https://claude.com/blog/claude-code-mods #AI #Anthropic #ClaudeCode #DevTools

Jarrod Watts turned waiting for Claude into multiplayer Doom. of course he did. His Intermission mod opens a match inside Claude Code and sends you back when Claude replies. macOS with Ghostty or kitty for now.

Claude Code 被“魔改”了。 Anthropic 上线 Mods：几行 TypeScript 就能改提示词、工具调用和界面，替换内置功能。还能让 Claude 自己生成并热更新。 这把 AI 编程工具的方向盘交给了用户。要小心：Mods 不进沙箱，权限和 Claude Code 一样大，别乱装 hhh #ClaudeCode #Anthropic #AICoding

day one of claude code mods and half of them are just things to do while claude works lol

@ClaudeDevs Just tried terminal-browser by zenbu-labs. Opened GitHub and browsed a PR right inside Claude Code 😅 I maintain a community mods list if you want more things to try: https://github.com/karanb192/awesome-claude-code-mods

@amorriscode @addyosmani @ClaudeDevs @claudeai @trq212 Install, in Claude Code: /plugin marketplace add tzafrir/whats-agent-doing /plugin install whats-agent-doing@tzafrir Free and Open Source: https://github.com/tzafrir/whats-agent-doing

You can save a LOT on AIs by making sure session cache is not outdated Here's the prompt to create tiny mod for #ClaudeAI Code that will show you remaining time: Build me a Claude Code mod (load the plugin-authoring skill first) called "cache-ttl" that estimates whether the main thread's prompt cache is still warm. - Hook `turn.step` (streaming generator): after `yield* next(e)`, for main-thread steps only (no `e.agentId`) that returned `usage`, save the time as the last cache refresh in `$.state`. - Estimate expiry as last refresh + TTL. Default the TTL to 1h, but make it a userConfig option ("1h" / "5m"). - Tick every 5s with `$.clock.every` (started in `session.start`) so the display updates. - Show it in the footer mode labels by hooking `ui.render` on `SessionMode`. Keep the existing `modes` (joined with " & ") and add a short colored label: - green "cache 58m" while warm - yellow in the last 5 minutes - red "cache outdated" once expired - Before the first request, pass through with `next(e)`. Validate it with `claude plugin validate` when done.

Your agent is in deep work and you have no idea what's going on. Install this free Claude Code mod (Desktop and Terminal) that says what each step is for, and when it's waiting on you 🤯 https://github.com/tzafrir/whats-agent-doing @amorriscode @addyosmani @ClaudeDevs @claudeai @trq212

thinkingアニメーションClaude mod

AI Daily · Oct 2⚙️ DEV🔹 Claude Code ships "mods" — customise behaviour & UI with a few lines of TypeScript. ⚠️ No sandbox, same local permissions as Claude Code itself.🔹 GitHub's gh-secure hardens public repos in one command — free for open source.🔹 Stitch by Google ships a CLI for driving local coding agents.🧠 MODELS🔹 Black Forest Labs releases FLUX 3 Image — multi-region editing, up to 10 reference images, native 4K.🔹 Microsoft's MAI-Transcribe-2-Streaming: 60 languages, first transcript in ~100ms. Preview — Microsoft says not for production.🔹 Perplexity open-sources a 27B decision model that outputs probability distributions, not text.🔹 Tavus's Griffin: 48% of testers thought it was human on a live video call, up from 2.4%.🏢 INDUSTRY🔹 California AG subpoenas OpenAI over AI-model cyber risk. FTC and several states running parallel actions.🔹 SoftBank completes its final $10B OpenAI tranche — $64.6B total for ~13%. OpenAI reportedly seeking $30B+ more.🔹 Tencent renting ~100k Oracle chips (~$7B, 5-year) — advanced AI chips it cannot get inside China.⚠️ OpenAI expanded GPT-6.1 Sol capacity — official claim ~2× launch-day speed, but early users report it's slower.

Claude Code has mods now, so I turned Claude into a tamagotchi 🥚 It lives in a pane beside your session and eats your work: edits bake cookies, passing tests bake cake, commits come as bento. Open source: /plugin marketplace add Syrul/claudagotchi

Claude Code just got mods: TypeScript functions that run before, after, or instead of anything the agent does. First one I'd write: block every git push until the tests pass. Catch: mods aren't sandboxed. Same access to your machine as Claude Code. Only install ones you trust.

Claude Code just got Mods, and most people are going to miss how big this is. A Mod changes Claude Code itself. What it shows you, what it's allowed to do, and what it does on its own. You don't need to know how to build one either. Claude can do it for you. Here are 10 ways to build one, and what I'd use each for: 1. A side pane. A live panel next to your chat. Use it for a dashboard of your running agents, today's to-do list, or test results that update as Claude works. 2. A band above the prompt. A thin row right where you type. Put your current goal there, or the checklist you always forget before you ship. 3. A status line entry. One small line that's always visible. Show cost so far, which branch you're on, or how long the task has been running. 4. Toasts. A quick pop-up when something happens. Get pinged when a long build finishes so you can stop watching it. 5. Guardrails on tool calls. Catch an action before it runs. Block pushes to main, block edits to your .env file, or require a second look before anything gets deleted. 6. Prompt rewrites. Change what you typed before Claude sees it. Auto-add your style guide, or turn your shorthand ("ship it") into the full steps you always mean. 7. Your own slash commands. Type /standup and get a summary of what this session got done. Type /client and load everything about the account you're working on. 8. Tools and agents Claude can call. Give Claude a new tool that talks to your CRM, or a "reviewer" agent it hands work to before calling a task done. 9. Background timers. Work that runs on a schedule inside your session. Check CI every 10 minutes, or wake Claude up when a new file lands in a folder. 10. Clean what gets saved. Rewrite the transcript as it's written. Strip API keys and customer names out before they ever hit your logs. I'm really excited about what's possible here. I'm testing out some helpful Mods right now, and I'm aiming to get a video out to show you more!

Made a Claude Code mod where Clawd acts out every spinner word 🦀 Beaming → beamed up by a UFO It uses none of your usage. Everything's animated locally, so your tokens go to the things that matter. @trq212 @ClaudeDevs #claude #claudemods

🚨 تحديث ضخم من Anthropic ​أعلنت Anthropic رسمياً عن ميزة Mods التي تتيح للمطورين إعادة تشكيل طريقة عمل Claude Code وتخصيص الواجهات (Custom UI) بسلاسة! ​📌 كيف تعمل ميزة الـ Mods؟ ​الـ Mods عبارة عن Hooks بسيطة مدمجة داخل Plugins بلغة JavaScript أو TypeScript تعمل داخل الجلسة وتتفاعل مع كل حدث فور وقوعه! ​يمكنك مشاركة تعديلاتك أو تحميل تعديلات المطورين الآخرين عبر Claude Directory! نحن نشهد الخطوات الأولى نحو عصر الواجهات المخصصة بالكامل.. مع قليل من الجهد والـ Tokens، أصبحت قادرًا على تشكيل البرمجيات على ذوقك الخاص! ​ما هو أول تعديل ستصممه لـ Claude Code؟ 🧵👇 https://x.com/ClaudeDevs/status/2105721434807083061/video/1

Yay :) I vibe coded a mod for vibe coders 😂 http://github.com/ruthannbravo/explain-it I kept finishing tasks in Claude Code like "cool… but what did you DO" 🥲 so now one button explains it, in plain English + the real technical terms. free + open source 🤍 fair warning: it's a yapper, but you decide when it explains (Explain It pop up appears when something has been done - you decide if you wanna use it). How it looks on terminal and on the Claude desktop:

Took the new @ClaudeDevs mods api for a spin and added a @linear band to the Claude desktop app. Inspired from the Github PR band, the irony being that I stopped reviewing code on Github entirely and review it now in Linear.

BREAKING: Anthropic launches "mods" for Claude Code, small TypeScript functions that can rewrite prompts, block tool calls and add new UI. For more, follow @theaideskio

universal-modder teaches Claude, Cursor, Gemini or any agent to reverse-engineer PC games, generate assets via fal, test mods in-engine, and record demos. loops through reconnaissance → route selection → code reading → building → generation → testing → knowledge storage. one toolkit, multiple agents.

Vždycky mě peklo, jak Claude Code píše dlouhé úryvky diffů na obrazovku. Nezajímají mne. Jenže někoho asi ano, takže tam pořád jsou. Tedy, už nejsou, díky Mods jsem si je odstranil. Modifikace jsou v Claude Code novinka. Přidal jsem si ještě počasí na kontextu. Jaká věc peče vás? https://www.vibecoding.cz/articles/claude-code/mody-pro-claude-code-anthropic-vas-pustil-pod-kapotu/

@0xPaulius Lmao, proof of concept Tbf don't know until you've tried

Claude Code Mods let you rewrite tool calls in-process. Easy to make the harness worse.

Claude Code mods are pretty cool :D

🔥 Claude Code теперь можно менять изнутри Anthropic представила Mods: небольшие функции на TypeScript, которые вмешиваются в работу Claude Code. Мод может изменить промпт до отправки модели, заблокировать или повторить вызов инструмента, скрыть секрет из ответа и добавить элементы интерфейса. Работает в CLI и настольном приложении. Моды распространяются через плагины. По сути, разработчик может настроить поведение самого агента под свой процесс, а не только давать ему инструкции. Устанавливать стоит из доверенных источников: мод выполняется с тем же доступом к компьютеру, что и Claude Code. https://claude.com/blog/claude-code-mods

طول ما كلود كود موجود كل يوم هتلاقي حد من التيم بتاعك فاتحله حوالي اتنين تلاتة pr على github، ولو التيم فيه ١٢ واحد فتبص تلاقي فيه +20 pr مفتوحين وفيه 40 pr اتعملهم merge خلال اسبوعين وطبعا انت محتاج تكون متابع كل التغييرات اللي بتحصل في الكود علشان خاطر تعرف لو فيه method اتشالت ولا اتغيرت، وتعرف هل ال changes اللي انت شغال عليها دي فيه حد ال pr بتاعته لامساها وهتعمل conflicts ولا لا، وهل اصلا الطريقة اللي انت شغال بيها لسه موجودة ولا طلعت pr غيرتها بطريقة تانية. وطبعا مفيش عقل بشري في وقتنا ده هيقدر يستحمل يقرأ فوق ال 40 pr كل اسبوع ويراجع ويفهم كل اللي حصل علشان كدا حاولت اساعد نفسي واساعد غيري وعملت plugin اسمها emsk-pr بتقول لل agent بتاعك ايه ال prs اللي مفتوحة دلوقتي وايه اللي اتعمله merge قبل ما يبدأ اي تعديل في الكود علشان يبقى عارف ايه اللي موجود وايه اللي اتغير وميرجعهوش تاني هي لسه تحت التعديل وكنت عايزة اجربها على scale اكبر من تجربتي الشخصية علشان كدا نزلت التويتة دي for testing Install in Claude Code: /plugin marketplace add 1AyaNabil1/emsk-pr /plugin install emsk-pr@emsk-pr Open source http://github.com/1AyaNabil1/emsk-pr

Claude Mods 这个方向太爽了。 一句提示，就能把 Claude 的工作方式和界面都改成你想要的样子。 每个人都该有专属的 Claude。

📙 Claude Code 모드 시작하기 (한국어 PDF) 이번 Anthropic 가이드를 한국어로 만들었어요. 23페이지 정도. https://drive.google.com/file/d/1i2xE1fQZvpz2-wB3vZZ37iNrkWwXbMNF/view 모드는 플러그인에 담겨 세션 안에서 실행되는 JS/TS hook. 이벤트를 관찰하고 수정하고 차단하고 커스텀 UI까지 그릴 수 있음. 이런건 Claude Code 팀에서 진짜 확장성 있게 잘 설계함.. 최고 👍🏻

Claude Code Modの最速日本語解説記事きてた。

I had to give #Claude mods a try, so created a stats line to see my limits