Community posts

Claude CodeにClaude Modsという機能が公開されたぞ！ Modsを使えば、Claude Code、ついに「使う」から“改造する”フェーズに入ったといえる。 Claude Mods / Function Hooks。 従来Hooksは外部プロセスを呼び出す仕組みだったが、Function HooksはTypeScriptをClaude Codeのエンジンに直接ロード。 これが面白い。 ・ツール実行前に危険操作を拒否 ・実行結果から秘密情報をマスク ・独自ツールを追加 ・UIを書き換える ・監査ログを差し込む ・組織ポリシーを強制する さらに prepend → user → append → builtin → core という5層構造で、個人Modより組織ポリシーを強くできる。 つまりこれは単なる拡張機能ではない。 AIエージェントに「EDRやプロキシのような統制点」を作る仕組みになり得る。 Claude Codeを企業利用するなら、 プロンプトをどう書くかより、この実行制御レイヤーをどう設計するかの方が重要になる。 まだearly access。 でもセキュリティ担当者は今のうちに読んでおいた方がいい。 https://zenn.dev/nogu66/articles/claude-code-function-hooks-claude-mods

Built a Claude Code mod using Jev that watches what you're asking and routes the turn to the right model. Quick renames go to Haiku. Real implementation? Opus. Architecture planning? Fable. @claudeai @ClaudeDevs @bcherny @lydiahallie @trq212

claude code modlarım artık status line'ı ve transcript'i kirletmiyor. sidebar: modların paylaştığı tek bir pane, iki katman. üstte duran durum, altta akan bulgular. en yeni üstte, pane dolunca en eski gider. api açık, her mod aynı yere yazar: await $.sidebar.set({ consumer: 'my-mod', key: 'src/users.ts', title: 'sql built from strings', lines: [{ text: 'src/users.ts:12', kind: 'warn' }], until: 'stream', }) until: 'session' üstte durur, until: 'stream' akışa düşer. set false dönerse sidebar kapalı demektir, modun kendi satırına geri döner. sidebar kurulu değilse çağrı hata verir, try/catch yeter. yani bağımlılık yok, modun tek başına da çalışır. button da koyabilirsin. basınca senin slash command'ını çalıştırır. ben kendi modlarımı bağladım. sen kendi modunu bağlayabilirsin. readme: https://github.com/kilimcininkoroglu/claude-code-mods/tree/main/plugins/sidebar

claude code sidebar mod sidebar: a single shared pane for mods, with two layers. the status stays at the top, while findings flow below it. newest items appear first, and when the pane fills up, the oldest ones are removed. the api is open, so every mod writes to the same place: await $.sidebar.set({ consumer: 'my-mod', key: 'src/users.ts', title: 'sql built from strings', lines: [{ text: 'src/users.ts:12', kind: 'warn' }], until: 'stream', }) until: 'session' stays at the top, while until: 'stream' goes into the flow. if set returns false, the sidebar is disabled, so the mod falls back to its own line. if the sidebar isn't installed, the call throws an error, so a try/catch is enough. in other words, there's no dependency, and the mod still works on its own. you can also add a button. when clicked, it runs your slash command. i've connected my own mods. you can connect yours. readme: https://github.com/kilimcininkoroglu/claude-code-mods/tree/main/plugins/sidebar

Claude Code compaction can stop summarizing. fast-jev-compaction: - Jev scores each tool call and result, stale ones dropped - what is kept stays verbatim, nothing rewritten - newest 6 pinned, threshold 0.5 - MIT, function hooks 2.1.274+, TypeSafe key https://github.com/tamaratran/fast-jev-compaction

TypeSafe shipped an agent skill for their System One API so Claude Code / other coding agents can design typed decision workflows directly. Install: claude plugin marketplace add typesafe-ai/skills claude plugin install typesafe@typesafe-ai Or for other agents: npx skills add typesafe-ai/skills --skill typesafe-ai Once installed, you just prompt naturally: "Use TypeSafe to route incoming support tickets by department, with human review for uncertain decisions." The agent handles designing the Choice/Score/Noul primitives, pulls current docs/cookbooks, and composes the typed judgments in code for you Nice example of the "skills as libraries" pattern the agent doesn't need the API memorized, it just knows how to go look it up and use it correctly

今日爆款 Skills🔥 1️⃣ Tencent/BrowserSkill（https://github.com/Tencent/BrowserSkill） 腾讯出品：让任意能跑 shell 的 Agent 使用你已经登录的真浏览器，Agent 窗口单独干活，验证码 / 弹窗可人工接力。今日热榜 +1306。 curl -fsSL https://raw.githubusercontent.com/Tencent/BrowserSkill/main/install.sh | sh “浏览器第一次是 Agent 的手，而且用的是你自己的 Cookie，而不是再开一只失忆的无头浏览器”！ 2️⃣ trailofbits/skills（https://github.com/trailofbits/skills） Trail of Bits 厂方安全技能仓：合约、Actions、C/C++、Rust、供应链、误报闸、差分审计、变异测试。不是一篇「请检查安全问题」。 /plugin marketplace add trailofbits/skills “安全技能第一次从审计公司仓库长出来，而不是社区口口相传的红队咒语”！ 3️⃣ google/skills（https://github.com/google/skills） Google 官方仓：面向自家产品与技术栈的 Agent Skills，厂方出厂源，不是 README 许愿清单。 npx skills add google/skills “云上技能第一次有第二家协议厂仓库，官方源不再只此一家”！ 4️⃣ EveryInc/compound-engineering-plugin（https://github.com/EveryInc/compound-engineering-plugin） Compound Engineering 官方插件：把复合工程系统接到 Claude Code、Codex、Cursor 同类工具。今日热榜在列。 按仓库 README 接入对应 harness “工程方法第一次按插件出货，而不是再贴一篇『请先写设计文档』”！ 5️⃣ Fission-AI/OpenSpec（https://github.com/Fission-AI/OpenSpec） 面向编码助手的规格驱动开发框架：先规格、再施工、可验证。今日热榜 +296。 按仓库文档接入你的 Agent “需求第一次变成可执行规格，而不是聊天记录里的愿望清单”！ // 总结 2026 的技能战争已经从「再写一条 prompt」变成「真浏览器工牌 + 审计公司宪法 + 厂方云技能 + 复合工程插件 + 规格驱动闸」。谁先把这五件套装进自己的 Agent，谁就把下一波生产力锁死在自己终端里。直接冲。

これは良さそう。利用料をかなり抑えられそう。 Claude Code用モデル振り分けMod「jev-model-router」 軽いタスクを確信度1.00でclaude-haiku-4-5に自動振り分け。tier/risky/レイテンシも丸見えで凄い。 ・typesafe API/Vercel AI Gatewayで動作 ・サブエージェント毎にモデルを自動選択 ・判定ログはワンクリックでコピー可能 ↓詳細

今週も直近1週間（9/13〜9/20）に note へ投稿された、Gemini・Claude・Notion AI のビジネス活用ネタを、無料のものだけに絞っていいね順に収集しました！🧐 経理・事務・営業・副業・業務自動化まわりの実践ネタ中心です 1位 〔Gemini／文章・資料〕X ♥約204 9/18 限界助教 ChatGPTで書いた文章をGeminiに自然な日本語へリライトさせ、その結果をGPT側で意味が変わっていないか照合する二段構えのスキルを公開。資料スライドの不自然な日本語つぶしにも使える。 https://x.com/genkAIjokyo/status/2101077468769857586 2位 〔Claude／業務自動化〕note ♥約189 9/15 Miccell(ミクセル) X投稿の予約までAIに任せる話。X公式MCPは従量課金なのに対し、Typefully MCP経由なら無料枠のまま下書き〜予約が通せる、という費用面の違いを整理している。 https://note.com/miccell/n/n65fe702a3927 3位 〔Claude／仕事〕note ♥約172 9/16 けい｜AIと共に未来を作るクリエイター Claude Codeを「エンジニアの道具」ではなく「仕事を任せる相手」として捉え直す入門。ターミナル・エージェント・Skillsといった用語でつまずく非エンジニア向けに、業務へ落とす順序をまとめている。 https://note.com/ak_hikiyose/n/ndc5e71261283 4位 〔Claude／副業〕note ♥約131 9/15 副業研究ラボ 受託業務を畳むつもりだったフリーランスが、Claude Codeで案件探し・個別提案・納品まで回せるようになった体験記。「ひとりの仕事」の採算がどう変わったかを具体的に語る。 https://note.com/brainy_yak6440/n/n3762860bb9fd 5位 〔Claude／コスト最適化〕X ♥約106 9/19 そら｜AgentSwarm 自動化オタク Claude Codeのモデル選びを自動化するMod「Jev Model Router」の紹介。依頼ごとに難易度とリスクを判定して、軽い作業は小さく安いモデル、危険な作業は深く考えるモデルへ振り分ける。 https://x.com/sora19ai/status/2101420066298536398 6位 〔Gemini／バックオフィス〕X ♥約98 9/18 しんご｜Copilotの専門家 会議の録音を渡すだけで、議事録を経由せずに稟議書がExcel形式で、しかも数式入りで出てくるというGemini Notebookの使い方。申請書づくりが多い部署ほど効く話。 https://x.com/shingo_copilot/status/2101063448880951586 7位 〔Claude／会計〕note ♥約82 9/13 岩瀬俊太｜税理士法人notte代表 会計事務所の記帳代行をClaude Codeに任せた実例。領収書や通帳データから帳簿を起こす、夜中までかかっていた作業がどこまで減ったかを代表自身が書いている。 https://note.com/bold_liger1034/n/na4eb84a701ef 8位 〔Claude／事務〕note ♥約81 9/16 さわきゆり｜インタビューアー・ライター ひとりで本業も事務も営業も抱えるフリーランスが、Claude CodeでAI秘書を立てて「業務チーム」を作り始めた話。孤独の解消という動機から入るのが独特。 https://note.com/teddy0914/n/ne10462e392f6 9位 〔Claude／営業〕note ♥約77 9/14 守友暁寛 / kubell 中小企業向けソリューションを扱う事業会社の営業が、セールス業務でClaude Codeをどう使っているかの社内実践レポート。開発ではなく営業現場での使い方に振り切っている。 https://note.com/akihiro_moritomo/n/n1d23943ad649 10位 〔Notion AI／業務効率化〕note ♥約22 9/15 ソウタ｜経営・マーケに使える最新AI解説 Notion AIの活用例21選。議事録要約やドキュメント草稿といった仕事別10例に加え、SNS運用への応用と、ChatGPT／Claudeとの使い分けまで整理した長尺記事。 https://note.com/ai__on/n/nd45cee47447b #AI活用 #ClaudeCode #GeminiNotebook #NotionAI #業務効率化

Steven Darlow acabou de soltar integração do Jev (TypeSafe AI) dentro do Hermes Agent. setup completo pra model routing, memory filtering, compaction selection, skill selection, computer use, browser use, private key-entry flow + dashboard de routing. pesquisa oficial (http://awesomejev.com, http://typesafe.ai, http://ai-sdk.dev): → o que é Jev (TypeSafe AI, lançado em early access 16/09/2026): modelo "System One" — NOVA categoria de frontier models pra decisões estruturadas rápidas que software consome direto. modelo atual `jev-1.13.0` (aliases: `jev-latest`, `jev-preview`). → diferente dos LLMs: Jev NÃO tem autoregressive loop, NÃO escreve texto, NÃO explica raciocínio. recebe um state (string/JSON/array) + set de typed questions, avalia TUDO em paralelo numa única chamada, retorna decisões estruturadas com probabilidades calibradas em 70-500ms. → não pode alucinar — saída tipada (não texto livre). otimizado pra outputs estruturados: classify, route, score, extract, branch. cai como "if-statement fuzzy" dentro de código comum. → pricing: $0.042 por milhão de input tokens. disponível via Vercel AI Gateway (`typesafe-ai/jev`) com `experimental_evaluate` da AI SDK 7 (a partir de 7.0.105) — provider nativo. → SDK oficial: `npm install @ai-sdk/typesafe-ai`. tipos primitivos: • `noul` (Vercel chamou de `boolean`) — true/false com probabilidade • `choice` — seleciona entre opções (até 255), retorna distribution completa • `score` — número em [0, n] • `typeSafeAi.evaluationModel('jev-latest')` é o entry point → exemplo típico de uso: ```js const result = await experimental_evaluate({ model: typeSafeAi.evaluationModel('jev-latest'), state: { message: 'I was charged twice. Please refund.' }, questions: { department: { type: 'choice', criteria: { billing: 'Charges, invoices, refunds', technical: 'Bugs, outages', other: 'Anything else' } }, severity: { type: 'score', criteria: ['Cosmetic', 'Workaround exists', 'Blocking'] }, requestsRefund: { type: 'boolean' } } }); ``` → use cases que mata: • agent guardrails — cada safety check que hoje é sub-agent caro vira call de 0.4s • model routing — decidir qual modelo chamar antes do call caro • intent routing — classificador rápido na frente do flow • retrieval re-ranking — re-rank BM25 com 1 Noul por query-passage pair • prompt injection detection — scoring de relevância antes do modelo principal • citation check — quoted citation supports the claim? → projetos do ecossistema: • typesafe-skill-router (DECRUX9812, ★3) — Hermes Agent: nomeia 1 skill pra carregar, antes do model call. opt-in, stdlib only, ~$0.001/turn • typesafe-mod (BeLazy167, ★2) — Claude Code mod: ranks installed skills per prompt + responde "this-or-that" do próprio agent quando confiante • ailerix (tylerjharden, ★1) — type-safe router. Jev banca cada request num typed catalog route • pi-jev-router (mejiasd3v, ★1) — automatic routing for Pi via Vercel AI Gateway • hermes-jev-router (ussyverse, ★0) — plugin experimental Hermes: Jev-assisted model routing plans com budget + capability constraints. API access pending • jev-router (prismhq, ★0) — open-source LLM router via Jev on top of LiteLLM • jev-skillful (bestagentkits, ★0) — per-prompt capability router: resolve skills, MCP servers, agents, commands, mede se injection ajuda • switchboard (anirudh-krovvidi, ★0) — guard pra LLMs em Jev com calibration/evaluation • ask-jev-skill (shantanugoel, ★0) — pra Hermes e outros agents • codex-router (0xNatoshi) — per-turn model & reasoning para Codex via System One → o que o setup do Steven Darlow destrava no Hermes: • model routing: cada turn escolhe o modelo certo • memory filtering: scoring de relevância pra injeção no contexto • compaction selection: decide o que manter quando contexto enche • skill selection: carrega só 1 skill relevante em vez de dump de skills • computer use: desktop automation roteado via Jev • browser use: web interaction routing • private key-entry flow: input seguro de chaves (zero-data-retention na Pro/Enterprise Vercel) • dashboard de routing: live observability das decisões → abordagem da `http://awesome-jev.com` é tratar Jev como classificador rápido na frente de flows caros. "expense only runs for requests that need them." → recurso crítico: agentes treinados em APIs LLM cometem 2 erros — fazem 1 question per call, inventam request fields. instalar a skill de agent da TypeSafe antes de pedir pra agent integrar Jev. → referência direta: `hermes-jev-router` por ussyverse (★0, status experimental). Steven sinalizou que pode ser útil pra @Teknium (Nous Research). #Jev #TypeSafe #Hermes #SystemOne #ModelRouting #Agentic #AI #LLM #DecisionModel #MCP #ClaudeCode #Codex #OpenCode #Pi #VercelAI #AISDK #DevTools #OpenSource #DevBR #BuildInPublic

AI関連の進化は凄まじいが、使いこなせている人はごく一部。Claude Codeが AGENTS. md のサポートMODを追加。Codex Astraのトークン消費削減を公式が提案。そして今回のJevもそうで、AIをX運用やnote販売に活かせてる人は少ない。

これから来そうなAI自動化GitHub TOP 5 今回は「Agentを増やす」より、 同じAgentを速く・安く・長く・壊れにくく動かす Repoが一気に伸びています。 GitTrendを軸に、特に気になった5つ👇 🥇 1. browser-use/jev-ultrafast https://github.com/browser-use/jev-ultrafast 9/18 GitTrend：+3.3k★級 超高速Browser Agent。 一般的なBrowser Agentは、 Screenshot ↓ Vision LLM ↓ 判断 ↓ 操作 を繰り返します。 Jev Ultrafastはページを操作可能な要素一覧に変換し、CLICK / TYPE / SELECT / SCROLLなどを選択。 文字入力が必要な場面を除き、LLMへの依存をかなり減らす設計です。 面白いのは、 「もっと賢いモデルを使う」ではなく 「LLMを呼ばなくていい場所では呼ばない」 方向に振っていること。 Browser自動化の速度・コスト改善という意味で注目です。 🥈 2. tamaratran/fast-jev-compaction https://github.com/tamaratran/fast-jev-compaction 9/18 GitTrend：+2.8k★級 Claude CodeのContext圧縮を変えるPlugin。 普通は古い履歴をLLMで要約しますが、これはTool Call / Resultを評価して、 重要 → 原文を残す 一部必要 → truncate 不要 → 削除 という方式。 つまり、 「全部うまく要約する」のではなく、不要なものを捨てる。 Agentを長時間動かすほど、 Context Windowの大きさ以上に 「何を忘れて、何を残すか」が重要になりそうです。 🥉 3. cloudflare/security-audit-skill https://github.com/cloudflare/security-audit-skill 9/18 GitTrend：+3.0k★級 CloudflareのCoding Agent向けSecurity Audit Skill。 特に面白いのが、 AgentがFindingを出す ↓ 別Agentが独立検証 ↓ deterministic validatorで確認 という構造。 ポイントはSecurityだけではなく、 「Agentの自己申告をそのまま信用しない」 こと。 AIに業務を任せるほど、実行Agentと検証を分離する設計は重要になりそうです。 latent-spaces/brag https://github.com/latent-spaces/brag 9/18 GitTrend：+1.6k★級 作ったProjectを、 1コマンドで短いLaunch VideoにするClaude Code Skill。 Project ↓ 構成 ↓ Demo素材 ↓ Short Video ↓ Share copy までつなげられます。 開発そのものだけでなく、 「作る → 発信する」までAgent workflowに入れる という方向が面白い。 OSSやWebサービスを作ってXで発信する人には、かなり分かりやすい用途です。 Tencent/BrowserSkill https://github.com/Tencent/BrowserSkill 9/18 GitTrend：+1.1k★級 Claude Code、Codex、Cursorなどから、 すでにログイン済みの実Browserを使える仕組み。 Agent専用Windowに加え、 既存Tabを明示的にborrow / returnする設計もあります。 ここで面白いのが、 BrowserSkill → Agentと実Browserを接続 Jev Ultrafast → Browser操作自体を高速化 という2つが同時に伸びていること。 今回の5つを並べると、今の流れがかなり見えます。 Context → fast-jev-compaction Browser接続 → BrowserSkill Browser実行 → Jev Ultrafast QA → security-audit-skill 制作・発信 → brag つまり次のAI自動化は、 「Agentをもっと増やす」だけではなく、 同じAgentを 安く・速く・長時間・壊れにくく動かす 方向がかなり重要になってきています。 個人的にまず追うなら、 Jev Ultrafastとfast-jev-compaction。 この2つは「モデルを賢くする」以外の方法で、Agent全体を改善しようとしているのが面白いです。

Claude Mod入門するなら、これを読んだらよさそう！ 実践した結果を元に、現状の仕様がまとまっている

Claude Mods 入門 | Claude Codeを自由にカスタマイズする https://search.app/fvYf8ZtwiuPR16659

新着ブログ更新のお知らせ 2026年9月20日版の技術ダイジェストを一挙紹介 判定特化AI Jev の高速・低コストな意思決定モデルから AIのテストマネジメント、セキュリティ事例、半導体市場動向、EmEditor の哲学、Claude Mods や Claude Code の新機能、OpenCodeLiteLLM の導… https://ift.tt/uyiqa9z

1/ Give Claude Code a .agt lookup in two commands: /plugin marketplace add ds1/agt-plugins /plugin install agt@agtnames The plugin bundles the read-only MCP server with a skill that teaches Claude when to use it.

やばい、Claude Codeの モデル選びを自動化するModきた。 Jev Model Routerは 依頼ごとに難易度とリスクを判定し、https://x.com/hityyhz/status/2101288260756484483/video/1 サブエージェント用モデルと 本体の推論effortを振り分ける。 ・軽い作業は小さく安く ・危険な作業は深く考える ・本体モデル変更は既定OFF 仕組みと注意点をリプで👇🧵

your coding agent can now call 100+ live tools straight from Claude Code - every call returns a verifiable receipt. `/plugin marketplace add https://zambo.dev/claude-plugin/marketplace.json` free. 20 calls/tool/day. no account. stop trusting "done" - verify it. https://zambo.dev

Something interesting happened in AI coding today: Claude Code finally supports AGENTS.md. Codex, Cursor, Gemini CLI and others already use it. But the bigger move is “mods”: Anthropic is opening parts of Claude Code’s agent harness for customization. Maybe the next AI coding battle isn’t: Which model is smarter? It’s: Who owns the agent runtime?

Your Claude Code install is missing a few things. Here are 40 additions (grouped by what they do): I've hoarded over a hundred Claude repos this year. These cover workflows, skills, memory and tools, plus reference libraries worth keeping open. BUILD A WORKFLOW 1. learn-claude-code → Learn how agents work. 2. karpathy-skills → Cut common coding mistakes. 3. superpowers → Plan, build and test. 4. ponytail → Keep the code simple. 5. gstack → Add planning and review workflows. 6. ECC → Add skills, memory and checks. 7. oh-my-claudecode → Coordinate agent teams. 8. Archon → Build repeatable coding workflows. ADD SKILLS 9. taste-skill → Improve your designs. 10. anthropics skills → Browse official skills. 11. mattpocock skills → Add engineering skills. 12. wshobson agents → Find specialist agents. 13. claude-plugins → Browse official plugins. 14. addyosmani skills → Add engineering checks. 15. ui-ux-pro-max → Get interface design guidance. 16. awesome-claude-skills → Find more skills. KEEP THE CONTEXT 17. planning-with-files → Keep plans in files. 18. claude-mem → Carry context across sessions. 19. codegraph → Map the code's connections. 20. graphify → Connect code and documents. 21. repomix → Pack a repo into one file. 22. agentmemory → Give agents persistent memory. 23. beads → Track work across sessions. CONNECT TOOLS 24. multica → Assign issues to coding agents. 25. firecrawl → Turn websites into usable text. 26. cc-switch → Manage your coding tool setup. 27. context7 → Fetch current code documentation. 28. vibe-kanban → Manage agent tasks on a board. 29. github-mcp → Work with GitHub from Claude. 30. playwright-mcp → Let Claude use a browser. 31. serena → Find and edit relevant code. 32. claude-code-router → Route model requests. 33. awesome-mcp-servers → Find more connectors. PROMPTS & USAGE 34. system-prompts-ai → Study AI tool prompts. 35. best-practice → Read Claude Code practices. 36. codex-plugin-cc → Add Codex reviews in Claude. 37. claude-hud → See your session usage. 38. rtk → Trim command output. 39. headroom → Compress what the model reads. 40. caveman → Get shorter replies (much shorter). So pick one for the job you're doing. Open its repo, follow the setup instructions and try it on a real task before adding the next one. My Claude resource vault → https://charliehills.substack.com/p/resource Repost ♻️ to help someone in your network.

claude code'a deneysel olarak gelen fonksiyonel hook olayı hoşuma gitti. CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1 claude şeklinde aktif edebiliyorsunuz. opus'a hazırlattığım limit-watch eklentisinin toplatığı istatistiklere göre haftalık limitim, limitlerin resetlenmesinden ortalama 1 gün önce bitecekmiş mesela. bunun yanında gemini'yi kullanan bir çok md hazırlattım. advisor, compact gibi konularda topu gemini'ye attırarak haftalık limit kullanımlarını rahatlatmaya çalışıyorum. benim hazırlattığım function hooks destekleri mod'ların olduğu repo: https://github.com/KilimcininKorOglu/claude-code-mods

GROK BUILD Sources: official http://docs.x.ai/build + http://x.ai/cli + your Grok Build bookmark folder (326 posts) Updated: 2026-09-19 NEW TODAY - Folder now 326 posts (4 new vs 322 on Sep 18) - Testers (blankspeaker): v1.0.38 · Sep 19 — long agent replies in prompt-suggestion transcripts no longer cut off; read_file on skill/instruction files configurable per deployment; pasted images survive yank/undo/history recall (notify if any cannot be sent); subagent overlays no longer stuck Cancelling; long quoted args wrap in permission prompts; tool-calling fixed via inline-only use_tool schema - Testers (blankspeaker / GrokInsider): v1.0.37 · Sep 18 — Mermaid flowcharts with complex labels/large sizes render and open correctly; `/context` legend percentages always add to 100%; `y` on a queue row copies the prompt; `/model` effort picker defaults to recommended; bash commands with quoted filename variables respect permission rules - Official changelog still Latest v1.0.34 · Sep 16, 2026 (checked Sep 19; v1.0.37–v1.0.38 not on http://x.ai/build/changelog yet) - Bookmarks also: KanikaBK on SpaceXAI persistent memory as a long-term collaborator (per-project Markdown notes + browse/organise) WHAT IT IS - A terminal-native coding agent with a fullscreen TUI - Also runs headlessly in scripts, or inside other apps via ACP - Understands the repo, edits files, runs shell commands, and manages long jobs - Open source harness: http://github.com/xai-org/grok-build - Different from Grok Bot (cloud teammates) and Digital Optimus (real-time video control) - Bookmark testers: native /remote-control support is coming (SpaceXAI confirmed); Warp now ships built-in Grok Build CLI with /remote-control (official grok Sep 11) INSTALL AND UPDATE [] macOS / Linux / WSL: curl -fsSL http://x.ai/cli/install.sh | bash [] Windows PowerShell: irm http://x.ai/cli/install.ps1 | iex [] Then: grok --version [] Update: grok update (stable) or grok update --alpha - Testers: v1.0.5 adds GROK_CONFIG env, auto worktree reclaim, hook-block wording, RTL, ACP effort, image/video call limits - Testers: v1.0.6 changes subagent spawning (breaking) and adds a customizable status line - Testers: v1.0.7 adds a first-class workflows catalog, Always/Never rules, and a timer-refresh status line - Testers: v1.0.8 improves MCP servers and workflows. Concurrent subagents start faster without freezing the parent or UI. Follow-ups send while a task is running. Ctrl+S stashes the prompt draft. /workflow autocompletes. MCP can request form input or URL consent. Workflow rows show current context usage - Testers: v1.0.9 adds agent budget + reasoning controls, plugin-provided agents, /feedback images, instant TUI mode switching, smarter clone/worktrees, resilient subagents, faster MCP startup, concurrent agents without rate-limit collapse - Testers: v1.0.10 — grok clone reuses matching local checkouts as linked worktrees for faster session creation - Testers: v1.0.11 — Auto/headless QoL: resume-picker headless sessions, configurable default permission mode, Auto subagent messages, mkdir/touch without prompts, no 10-hour monitor timeout, faster background waits, turn duration after /resume - Testers: v1.0.12 — faster worktrees; MCP retry; subagent wait after interjection; hook descriptions; context/token accuracy; auto-compaction bar; .NET watcher fix; Auto recap during background streams - Testers/Official: v1.0.13 · Aug 28 — compressed CLI downloads; truncated responses continue; hooks can confirm (not only allow/deny); inference retry on stalls/drops/5xx; Windows ~/.grok and worktrees; more reliable session save; faster MCP startup when auth configured; truncated tool calls with complete args execute; large-image session fix; scheduled-task UUID IDs + stop reminder - Testers: v1.0.14 — PostToolUse hooks feed model feedback; grok usage <session-id> per-turn tokens/cost; proactive OIDC refresh; stronger subagents; ~70% smaller Windows CLI downloads; large-session memory no longer blocks turns - Testers: v1.0.15 — background model connect/token refresh/MCP at startup (faster first reply + new sessions); memory consolidation deferred off exit path; /copy and /export tip after repeated drag-copies - Testers: v1.0.16 — MCP servers at session bind for workspace integrations; enterprise model restrict via signed requirements.toml; 1h default wait ceiling for long subagent/task output; MCP OAuth from /mcps deadlock fix; sessions recover after expired-token network interrupts - Testers: v1.0.17 — MCP multi-round-trip elicitation (input_required → resume); quieter next-prompt ghost suggestions; /btw panel table cell select/copy as TSV - Testers: v1.0.18 — managed policy blocks disallowed MCP servers and marketplace installs before write; interactive Chart.js charts; richer image-gen params via composer state; configurable per-model rate-limit retry; per-model mTLS for secure upstream; faster startup/first responses with more work in background; session/subagent/auth/composer/UI fixes - Testers: v1.0.19 — scheduled /loop tasks always run in background (no turn inject); session resume reports running loops/subagents/workflows; headless `grok -p --worktree`; dashboard shows new sessions immediately + /usage from dashboard; transparent terminal theme; faster first prompt after login with MCP; UI freeze/crash fixes - Testers: v1.0.22 — first-party desktop MCP; `grok-workspaced daemon` for Computer Hub folder expose; continue finished subagents; background agents return real output; real line-number diffs + auto-expand on edit permission; built-in tools beat user MCP name collisions; remote-control device grouping; Auto blocks destructive `git checkout --`; faster open/resume (MCP config off critical path) - Testers/Official: v1.0.23 — bare URLs and emails inside table cells stay clickable on every wrapped line - Testers/Official: v1.0.24 — Esc no longer cancels a running turn (use Ctrl+C) - Testers/Official: v1.0.25 · Sep 9 — `/workflow` pause/stop for agent-launched runs; durable background-task snapshot across reconnect; silent successful hooks; full Bash pager output; `/theme` alias match; voice inserts at cursor; scheduled-task cleanup reminders; `/feedback` immediate send; dashboard/actions/UI fixes - Testers: v1.0.26 · Sep 10 — subagent interjects interrupt long background waits; scrollback labels how messages were sent; feedback form available in minimal mode - Testers: v1.0.27 · Sep 10 — queued messages no longer duplicate when sent immediately; dashboard spinners stay live during background tasks - Testers: v1.0.28 · Sep 10 — mid-message `/btw`; per-model `reasoning_summary` in config.toml; safer multi-process auto-update; soft-wrapped prompt Up/Down fix - Testers: v1.0.29 · Sep 11 — pager Tab cycles prompt/scrollback/whole dock; Ctrl+G hide/show dock; concurrent `grok agent stdio` same GROK_HOME startup crash fixed; MCP 2026-07-28 servers; Windows Client-ProjFS-off startup fix; idle dashboard CPU paint reduced - Testers: v1.0.30 · Sep 11 — session header single-row (overlay context + Dashboard); tmux pane lag fixed; elapsed time into hours; Watchers/Loops next-run due; workflow status clickable in dock - Testers: v1.0.31 · Sep 13 — multi-agent readability (running vs finished subagent group counts); dashboard search/resume predictability; subagent status labels; dock sections/paths; cancellation banners/messages; MCP long server-prefix tools fixed - Testers: v1.0.32 · Sep 14 — plugin/skill listings reflect latest config immediately (even before a session); Windows ARM64 first-handshake TLS crash fixed; slash commands like /feedback typed during plan approval execute correctly - Official/Testers: v1.0.33 · Sep 15 — MCP tool results include structured JSON when the server provides it; cancelled tool calls tell the server to stop; manage memory from /memory (delete with two-press x); GROK_GROVE=1 enables clone + worktree when knobs are set; long-running session + agent reliability; SSH Alt+Enter footer; macOS paste image fix; and more (see changelog) - Official/Testers: v1.0.34 · Sep 16 — Memory generally available (remembers conventions, decisions, project facts across sessions; /memory browse; /dream organize notes into topics); Markdown headings receive theme colors correctly - Testers: v1.0.35 · Sep 17 — better MCP status reporting and tool rendering; Swift syntax highlighting; cleaner copy/paste in minimal mode; improved /memory UI and much faster memory - Testers: v1.0.37 · Sep 18 — Mermaid complex labels/large sizes render correctly; `/context` legend always sums to 100%; `y` on queue copies prompt; `/model` effort defaults to recommended; bash quoted filename vars respect permission rules - Testers: v1.0.38 · Sep 19 — long agent replies in prompt-suggestion transcripts no longer cut off; read_file on skill/instruction files configurable per deployment; pasted images survive yank/undo/history (notify if unsendable); subagent Cancelling overlay fix; long quoted args wrap; inline-only use_tool schema - Official changelog Latest v1.0.34 · Sep 16, 2026 (checked Sep 19; Memory GA + v1.0.31–v1.0.33; testers already on v1.0.35–v1.0.38 not listed yet) - Bookmark testers: Grok Build mode is now on all SuperGrok and X Premium plans, web and mobile - Testers: official Netlify plugin for build and deploy - Official/Warp: Grok Build CLI runs natively inside Warp (rich prompts, /remote-control, file explorer + code review) - Testers: Quo communications plugin — SMS (individual/group/bulk), contacts, tasks, message/call transcripts, missed-call follow-up (`grok plugin install quo --trust`) - Bookmark testers: Grok Build can find/sequence footage and do real image/video editing (cut, assemble, Photoshop-style changes) from a brief — not only generate; save preferred designs/workflows as skills - Bookmark testers: Grok Build Mode can ship simple web apps (example: image watermarker) - Bookmark testers: Grok Build can drive Blender 3D workflows (examples: rocket launch site from scratch; Blender + Imagine car-chase cinematic clip) - Bookmark testers: a usage-reset control can set usage back to 0% - Changelog: http://x.ai/build/changelog - Evaluate Grok 4.6 with the Grok Build harness (http://grok.com/build) - Docs: http://docs.x.ai/build/overview - Download page: http://x.ai/cli SIGN IN [] First launch opens a browser for authentication [] Headless or remote: grok login --device-auth [] Or: export XAI_API_KEY="xai-..." then grok [] Sign out: grok logout [] In TUI: /login and /logout START A SESSION [] cd your-project then grok [] First prompts: Explain this repo. or src/main.rs Walk me through this file. [] grok -p "Explain this codebase" for a one-shot headless run [] grok -c continues the most recent session in this directory [] grok -r resumes a session by id (or the most recent) [] grok inspect shows discovered config, skills, plugins, hooks, and MCP [] grok models lists available models [] Switch model in TUI with /model LEARN THE TOOL [] Run /tutorial or /help [] /release-notes or /changelog for the current version [] Read AGENTS.md in the repo (also CLAUDE.md if you are migrating) [] Import Claude settings with /import-claude [] Carry over rules, skills, and MCP from Claude / Cursor / Codex [] Keyboard shortcuts: /help and Ctrl+. [] Voice: Ctrl+Space or /voice [] Shift+Tab cycles Plan, Auto, Always-approve HOW TO WRITE A TASK [] Outcome: what should be finished [] Sources: files, folders, URLs, or mentions [] Constraints: what not to touch, and when to ask [] Deliverable: diff, report, test, or published app [] Review point: plan first, or stop before commit/push [] For big work: /plan first, then approve [] For long ownership: /goal [] For research first: /deep-research [] For a sick laptop: tell it what is broken and let it investigate [] Testers (tetsuo): start with work you can check — name the behavior, give a reproduction command, limit files, list final checks, say where to stop [] Testers (tetsuo): inspect first (git status --short, grok inspect); give executable feedback (compiler/linter/tests) so the agent can loop; review the full diff and the command behind every green check CORE COMMANDS [] /new starts a new session [] /home returns to the welcome screen [] /resume and /sessions switch sessions [] /fork branches the current session [] /rename or /title names the session [] /session-info shows session details (click or drag to copy) [] /context shows token and skill/MCP cost [] /usage shows credits [] /memory browses session/project memories [] /dream organizes recent notes into topics [] /compact condenses history [] /rewind truncates conversation history (confirms; does not wipe files) [] /export saves the conversation [] /btw asks a side question without derailing the turn (v1.0.28: also works mid-message — text after `/btw` is the side question and stays out of the main transcript) [] /check-work has a subagent review the code [] /feedback files a report [] /doctor and grok doctor fix terminal, tmux, clipboard, keyboard [] grok du shows disk use under ~/.grok [] grok wrap runs a command with local clipboard support [] grok trace exports debug traces [] /loop runs a prompt on an interval (auto-expires after 7 days) [] Queued follow-ups can send as interjections if you turn that setting on [] Ctrl+4 opens the Agent Dashboard [] Ctrl+U can continue a Claude Code session inside Grok Build PLAN MODE AND PERMISSIONS [] /plan or Shift+Tab enters plan mode [] Only the plan file can be edited until you approve [] Auto and always-approve do not skip plan review [] a approves, s requests changes, c comments, q quits plan [] /view-plan reopens the preview [] /auto auto-approves safer tools [] /always-approve skips most prompts (deny rules and hooks still apply) [] Edit always-allow bash as free-form globs [] Restrict web search domains with [toolset.web_search] in config.toml [] PreToolUse hooks can rewrite tool input, not only allow or deny [] StopCancelled reports when a turn is interrupted or rejected [] Pin auth to API key or OIDC in config.toml if needed PARALLEL WORK AND AGENTS [] Agent Dashboard: /dashboard or Ctrl+4 [] Run feature, review, fix, and test agents from one place [] Pause, check progress, and jump into any agent [] Worktrees: grok -w or Ctrl+W (isolated git copy under ~/.grok/worktrees) [] grok worktree list / show / rm / gc [] /fork --worktree forks into a worktree [] /goal owns a long engineering objective (status, pause, resume, clear) [] /create-workflow authors a repeatable multi-agent pipeline [] /workflow launches, pauses, resumes, or stops a run [] /workflows is the live run dashboard [] /deep-research starts a built-in research workflow [] Tools and MCP servers receive GROK_SESSION_ID [] Background tasks and TODOs survive compaction VOICE AND MEDIA [] Dictate with Ctrl+Space [] /imagine generates an image [] /imagine-video generates a video [] Ask it to boost, compress, and re-encode video audio in plain English [] One-prompt 3D: ask it to model in Blender and watch it drive the app [] Morning voice-memo automations can research and read you a briefing [] Disable image/video tools via config or env if you do not want them [] Testers: save preferred image/video edit designs and workflows as skills so repeats match your style MEMORY (v1.0.34 GA) [] Grok Build remembers conventions, decisions, and project facts across sessions [] Notes capture in the background after useful work [] /memory to browse and manage memories (delete with two-press x) [] /dream to organize recent notes into topics [] The more you build with it, the more useful it becomes (official grok) [] Bookmark testers: persistent memory feels like a long-term collaborator — per-project Markdown notes + global prefs (KanikaBK / SpaceXAI) SKILLS PLUGINS AND MCP [] /marketplace installs plugins [] /plugins /skills /hooks /mcps open the same extensions modal [] Skills live in ./.grok/skills, ~/.grok/skills, and plugin skill folders [] User skills appear as slash commands; use /local:name if names collide [] grok plugin install / enable / disable [] grok mcp list / add / doctor [] Type / to reference a skill [] Testers: save preferred designs and editing workflows as skills so Grok Build reuses your style every time [] Common marketplace plugins: Exa, TinyFish, MongoDB, Vercel, Sentry, Cloudflare, Chrome DevTools, Supabase, Neon, Firecrawl, Figma, Railway, Stripe [] Official/Testers: browser-use plugin — /plugins or `grok plugin install browser-use --trust`; enable Chrome remote debugging to drive your signed-in local Chrome (or isolated cloud browser; local uvx, no API key for local Chrome) [] Managed MCP servers come from the gateway catalog [] Add a local directory as a marketplace source if you keep private plugins [] Testers (v1.0.33): MCP tools can return structured JSON when the server provides it; cancelled tool calls notify the server to stop [] Official (v1.0.34): /memory browse/manage; /dream organize notes; Memory GA across sessions [] Testers (v1.0.33): GROK_GROVE=1 enables both clone and worktree features when specific knobs are set WORKTREES AND FILES [] Worktrees start from current HEAD, including uncommitted changes [] grok -w --ref main starts from a clean ref [] Ending a session does not delete the worktree; run grok worktree gc [] Keep durable files in the repo or ~/.grok, not temp dirs [] Config: ~/.grok/config.toml (Windows: %USERPROFILE%\.grok\config.toml) BUILD MODE AND PUBLISHING [] Build Mode on http://grok.com (and mobile where available) [] One prompt can become an app on a http://grok.me domain [] Export source to GitHub [] Examples: http://driver.grok.me, http://bbox.grok.me, http://gravity.grok.me [] SuperGrok Heavy may be required for some publish features [] Bookmark testers: Grok Build is now available for free users too — describe what you want, Grok builds apps/websites/games/dashboards, then publish and share [] Testers: Tesla Owners Silicon Valley — creating from a prompt is open to free users; a live http://grok.me URL still requires a paid plan [] Testers: Android (X Freeze) — GitHub push, secrets store, invite-only lock, download builds, custom domains, share to X; end-to-end ship from a phone [] Testers: Android Library (Menu > Library) — Imagine generations, Grok Build mode apps, and files created/uploaded; mobile access like desktop HEADLESS AND AUTOMATION [] grok -p "task" for scripts [] grok -p "task" --output-format streaming-json [] grok agent stdio for ACP in other apps [] grok serve --protocol acp --port 9120 for a local ACP server [] Custom models go in ~/.grok/config.toml under [http://model.name] COMMUNITY TOOLS [] GrokTerm: multi-harness terminal host with voice [] desktop-harness: Mac Accessibility control (http://github.com/xfreeze2/desktop-harness) [] Quill: universal SuperGrok voice layer on macOS [] AFK Pilot / Grok Build Desktop: remote edit and syntax highlight [] Grok Build for VS Code (Community) extension — Testers: Routines dashboard for scheduled executions across Grok Build, Codex, and Claude Code (89K+ installs; not affiliated with SpaceXAI) [] Unity CLI / MCP for game work SAFETY AND USAGE [] Prefer plan mode before large edits [] Prefer always-approve only after you trust sandbox and deny rules [] Opt in to share traces only if you want to (off by default) [] Privacy banner can be dismissed from Settings [] Video tools may require ZDR storage [] Check /usage before long multi-agent days [] Grok 4.6 may run double-token promos; check current offers OFFICIAL LINKS [] Overview: http://docs.x.ai/build/overview [] Modes and commands: http://docs.x.ai/build/modes-and-commands [] Plan mode: http://docs.x.ai/build/features/plan-mode [] Worktrees: http://docs.x.ai/build/features/worktrees [] Skills and plugins: http://docs.x.ai/build/features/skills-plugins-marketplaces [] CLI reference: http://docs.x.ai/build/cli/reference [] Changelog: http://x.ai/build/changelog [] Install: http://x.ai/cli [] Source: http://github.com/xai-org/grok-build

Claude Code just got a lazy senior dev built in Meet ponytail — an open source plugin for Claude Code that behaves like the most experienced (and laziest) dev on your team, in the best way possible. Instead of generating a fresh generic app from scratch every time, it actually checks your existing codebase and templates first, matches your style, and builds off what's already there. No cap, it barely touches anything it doesn't absolutely have to. Install is one line straight in your terminal: /plugin install ponytail@ponytail And you're cooking. The best part: it works with any AI model, not just Claude, and you only pay Anthropic for the actual usage — no stack of extra subscriptions on top. Lowkey the most efficient setup I've seen for shipping real projects instead of just vibes-coding another throwaway demo If you're tired of AI tools rewriting your whole app every time you ask for one small change, this fixes that, straight up

INTRODUCING JEV MODEL ROUTER FOR CLAUDE CODE This Claude Code Mod lets you run Jev through the direct @typesafeai API or @vercel AI Gateway. On every request you send to Claude Code, Jev picks: → the subagent model → the main model (only at session start, so the cache doesn’t break) → the effort level Install with one command: npx claude-code-templates@latest –mod productivity/jev-model-router Full Mod details: https://www.aitmpl.com/component/mod/productivity/jev-model-router

fast-jev-compaction is a Claude Code plugin that swaps lossy compaction summaries for Jev keep or drop decisions. Tool calls stay verbatim when they still matter. 3,763 stars. https://github.com/tamaratran/fast-jev-compaction

jev is insane!!😳 i've been using it for the past 6 hours and heres how i would recommend to use it to 100x your workflow. quick context: jev is a model that can't write text. it only decides. you send it state and questions, it answers all of them in one pass, 70-500ms, and output is free. here's how to actually use it. the whole api is 3 question types: 1. choice — pick one from a list (up to 255 options) 2. score — rate on a 2-10 level scale 3. noul — yes/no, returns a 0-1 probability one endpoint: POST https://api.typesafe.ai/v1/systemone setup: claude code → claude plugin marketplace add typesafe-ai/skills elsewhere → npx skills add typesafe-ai/skills --skill typesafe-ai then just tell your agent: "use the typesafe skill" (or skip it: pip install typesafe-sdk) a real example. support ticket: from typesafe_sdk import Choice, Noul, Score, TypeSafeClient client = TypeSafeClient() response = client.system_one( state={"ticket": {"subject": "duplicate charge", "messages": [...]}}, questions={ "department": Choice(instructions="which team should handle this", criteria={...}), "is_urgent": Noul(instructions="the message conveys urgency"), "frustration": Score(instructions="how frustrated the customer appears", criteria=[...]) } ) it answers all three in parallel. one call, ~200ms. the part that matters: batch everything. adding questions barely moves the latency, and output is free. ask every question you might need, even ones you'll throw away. gate on confidence, not the answer. under 0.5 escalate to a big model or a human. 0.85+ before anything irreversible. never let it invent options. build the candidate list in code, then let it pick. it's a decider, not a writer. don't ask for explanations. the honest part: text only. no images, no audio. on broad benchmarks it loses to frontier models — ~68% on TypeSafe's own 4-workflow test, about mid-tier. it's the wrong tool for chat, code gen, or anything that needs a written answer. it wins on narrow, well specified decisions. which is most of what your agent does all day. and those numbers are TypeSafe's own, self-run, not independently verified yet. early access is waitlisted. many people got in thesame day http://typesafe.ai

🎬 Powermove is out: a free, open-source video editor (macOS) with a tiny kernel — timeline, effects and UI all live as hot-reloadable mods you can ask the agent (Codex/Claude Code) to write, per site. Real layers and keyframes, undoable agent edits, export to video, web or SVG. Live on Product Hunt today. https://trypowermove.com

sovereign-skills v6.5.12 — 20 Claude Code skills for the whole project lifecycle, from setup to governance. New: the adoption gate now vets plugin hooks & MCP tool descriptions; handoffs tag what was verified vs. only discussed. MIT · Codex-compatible https://github.com/AlexZio00/sovereign-skills

【速報｜Claude Code v2.1.277 リリース】 Claude Codeが「AGENTS.md（他のAIツールと共通の指示書ファイル）」を読めるようになった！ 発表したのはClaude Code開発チームのThariq氏。投稿は1日で2.5万いいね😳 これ、地味に見えて 複数のAIツールやチームで開発する人の前提が変わるアップデートです。 どこが変わるのか、公式の変更履歴とGitHubの実装から要点だけ抜きます。 これまで 「Claude Codeの指示書＝CLAUDE.mdだけ」 でしたが── その前提、崩れました。 【今回のアップデートで何が変わった？】 フォルダにCLAUDE.mdが無ければ、代わりにAGENTS.mdを読みに行く。 Codexなど他のAIツール向けに書いたAGENTS.mdを、Claude Codeもそのまま指示として読む。 つまり👇 ・CLAUDE.mdが無いプロジェクトでも、AGENTS.mdの指示がそのまま使われる ・/config の「Project instructions」で、どの指示書を読むかを4択から選べる ・両方読む設定（claude-md-and-agents-md）も選べる ・この機能は「Claude Code mods」という拡張の仕組みで作られていて、コードはGitHubで公開中 クラウド各社経由（Bedrock・Vertex・Foundry）で使っている人はまだ非対応。 Claude Codeを 「Codexと併用している人」ほど、 今回のアップデートは確実に体感できるやつです✅ 指示書を1本にまとめるなら、Anthropic社内でCLAUDE.mdがどう書かれ、どう使われているかを先に知っておくと迷わない。 Anthropic社内の運用ベストプラクティスを全部まとめた記事はこちら👇

家人们好消息！Claude Code 终于开始支持 AGENTS.md 了！！！ 刚刚 @trq212 宣布， 从 2.1.277 版本起，默认情况下，如果工作目录及其上级目录里没有 CLAUDE.md 等项目指令文件，Claude Code 就会转而读取 AGENTS.md。 这个行为可以在 /config 中调整。 以前要让 Claude Code 也用这份说明，通常得在 CLAUDE.md 里引用它，或者建立软链接。 现在，已经采用 AGENTS.md 的项目，可以少做一层适配。 更值得关注的是 Thariq 后面补充的消息：这次支持 AGENTS.md，是通过一个内置 Mod 实现的。 他们正在准备 Claude Code Mods，用来让开发者自定义 Claude Code 的 Harness。 按照他的说法，未来开发者也能按自己的需求，构建自定义的项目指令模块。 但我我更关注后面这一步。 支持通用文件，能减少切换工具的麻烦；允许定制指令模块，则有机会让团队决定自己的项目规则应该怎样接入 Agent。 工具可以各有各的能力，项目规则没必要各写一份。以后选哪个 Agent，更多看它干活怎么样，少看它认哪份说明书。 PS：具体能改到什么程度，还要看后续开放的接口。