DEV TRENDS / CATEGORY
2026年10月21日 – 2026年10月27日 内首次发现的 95 个新项目，已跨 5 个来源去重。
2026年8月5日 – 2026年10月27日 内首次发现 972 个项目
Paste your git log; the grader lists every file two sessions (two authors, or two branches) changed less than an hour apart, ranked by how often it happened. Those are the edits most likely to undo each other when Claude Code, Codex or a teammate work in parallel. It runs entirely in your browser: nothing is uploaded, no signup. Try the sample log to see a report in one click. A README badge shows the count for a public GitHub repository.
原文Spacesheep (YC F26) turns any agent's work into a space: an interactive doc, live dashboard, mockup or app. Share it with specific people, your team or publicly, and collaborate on it together.
原文按用量计费的 AI 模型网关，一个 Key 双协议通用：Claude Code 设置 ANTHROPIC_BASE_URL、OpenAI SDK 改 base_url 即可接入，价格公开、每笔请求成本可查，支持组织内多发 Key 与用量审计、多上游自动故障转移；名字与 logo 源自 HTTP 彩蛋 418 I'm a teapot
Token Kingdom turns your Claude Code usage into a cozy 3D village. Every token you spend becomes gold, wood and stone. Build, fish, cook, defend against raids and visit your friends' kingdoms. One plugin install. Free.
原文Nekot shows every Claude Code and Codex usage window on your iPhone, when each one resets, and alerts you at 80% and 95% so you never hit the limit mid-task. Your tokens stay in the Keychain: no Nekot server, no account.
原文Pling is a real SSH, Mosh and SFTP terminal for iPhone, iPad and Android, built for working with coding agents. Jump back into tmux or Zellij sessions, approve Claude Code's risky commands from a push notification, or describe a task and let the Command Agent run it. Plus push alerts from any script with one HTTP POST, live server monitoring, remote desktop and end-to-end encrypted sync. Free to start; Pling Pro unlocks the rest.
原文把所有 AI 编码智能体收进同一个地方 —— 并让它们协同工作。 亮点 可以让任何兼容ACP的智能体在一个界面协同工作，支持桌面、服务器、Docker 与移动端，也能通过 Telegram、飞书和微信与智能体对话。 截图或演示视频
Kinaya is a keyboard-first browser for macOS. Ads and trackers are blocked inside the engine (Brave's adblock-rust with EasyList and EasyPrivacy) before a request leaves, and there is no telemetry. A built-in MCP server lets Claude Code read pages, click, type and take in its own background tab, using your real logins. Also: command palette, private and Tor tabs, encrypted password vault, reader mode. $19 once, 14-day free trial.
原文Session and weekly limits for Claude Code, Codex, Cursor and Grok, with a pace check that tells you if you'll make it to the reset.
原文See every Claude Code session on one live screen: which one is working, which needs you, what its agents are doing and what it costs. Each chat runs in its own window and keeps going in the background. Free and open source.
原文Graphicle is a Rust port of Graphify's open-source code-graph engine, 3 to 22.7x faster than Graphify 0.9.91 in all 39 of our measurements, plus project memory: docs, decisions and past agent sessions, cited to file and line over MCP. Local, Apache-2.0.
原文TraxAgent is an AI spend intelligence platform: it turns your Claude and OpenAI APIs, coding agents like Claude Code, Codex and Cursor, and Claude and ChatGPT AI usage from across your domains into actionable insights and consolodated spend. Risks flagged early, costs cut, opportunities found, each with the exact fix for optimization. Read-only, no prompts stored.
原文Turn your Claude Code and Codex sessions into a living fantasy village. See agent activity at a glance, then click a hero to inspect context usage, skills, plugins and subagents. Agent Quest reads existing local session logs while you keep using your usual tools. Keep multiple sessions in sight on a second screen. Customize the village and export Markdown or JSON reports for review. Free, MIT-licensed and open source. Runs locally in your browser with Bun.
原文一个 Windows 无边框、无任务栏图标的 AI 订阅额度 HUD 悬浮窗与副屏仪表盘，同屏实时监控 Claude Code、ChatGPT Codex、Grok、Google Antigravity、LiteLLM 等多账号的 5 小时与 7 天限额窗口；支持一键探测并兑换官方隐藏的限额重置券；双击 Ctrl 秒切 mpv 动态壁纸。 解决了什么痛点？ 1. 多账号与多平台额度盲盒：做严肃编码任务时，多开 Claude Code / Codex / Grok / agy 经常突然撞上 5 小时限额墙或周上限。以往要在各个网页或 CLI 里繁琐查询，usage-float 直接以极低资源占用在副屏或桌面角常驻展示全部账号的动态进度条与剩余重置时间。 2. 官方隐藏福利一键兑换： Anthropic 为 Claude Code CLI 用户提供了周重置券（cedarember）与 5 小时碰墙重置券（junipertide）； OpenAI 官方对部分 ChatGPT / Codex 订阅也发放了限额重置券（/backend-api/wham/rate-limit-reset-credits）； usage-float 模拟官方 CLI 原生协议安全嗅探重置券资格，有券时直接在界面提示，点击即可一键兑换回满血，绝不浪费。 3. 无感常驻与桌面摸鱼/护眼切换：副屏平时是干净专业的黑金/暗黑仪表盘，双击 Ctrl（或自定义热键）无缝切换为你本地收藏的 mpv 4K/60fps 动态壁纸或画廊，兼顾生产力与视觉愉悦。 4. Agent 本地查询友好：除了图形界面，提供 python usagefloat.py once（或读取本地原子写入的 ~/.usage-float/usage-cache.json），方便本地 Agent 路由器无网络请求开销直接获取账号剩余额度。 配套开源工具自荐（同一开发者生态） 为了让多订阅 AI 编码体验更完整，我们在同一生态下还开源了另外两个轻量实用的工具，供感兴趣的同学参考： 1. AgentCompass
A calm Mac app for running Claude Code sessions side by side. Perch keeps every session, PR, and link for a piece of work in one place, and tells you when something needs you.
原文ssh zydra.sh There's no sign-up or anything, and you get a 15-minute of a graphical Linux VM drawn entirely in your terminal. It has a browser, an editor, and it runs Doom. It works best in Kitty or Ghostty. WezTerm and Konsole also work, as well as Sixel and iTerm2 terminals, though with less detail. Text-only terminals support non-graphical apps: shells, editors, and other TUIs. Why? For a few months I ran coding agents in tmux on a VPS over SSH, and it sucked. I couldn't paste images, scrolling broke and I didn't wanna figure out why, I had no notifications for when an agent finished, I couldn't see browsers agents opened to smoke-test, logins drifted between my laptop and the server, I couldn't check in from my phone, and to know what an agent had done five minutes ago I had to read through opaque chunks of tool calls. These issues added up and had me in a limbo: They made my setup bad enough to annoy me every day, but not quite bad enough to stop my real work and fix it. How it works: Zydra is a custom (Smithay-based) Wayland compositor in a microVM with an AMD GPU, implementing a series of terminal protocols that solve many issues but are often ignored: Kitty graphics for windows, OSC 52 for the clipboard, images included, OSC 9/99/777 for notifications, SGR mouse for scrolling and clicks. VNC is limited to sending pictures of the screen and diffing frames to find what changed. In Zydra, apps directly tell the compositor what changed, and terminal text is sent as text you can select and copy natively. The browser is a WebKit build that delivers frames to the compositor as DMA-BUFs without a readback, and agents drive it through Playwright. Alt+P is sort of like Raycast: browse apps, linked GitHub, websites, and SSH targets from it. You can watch the desktop at any point in time and rewind to periodic snapshots. To achieve this, the desktop records into a timeline file, a kind of structured video, and the VM takes snapshots along it. You can traverse back like a livestream to see what happened while you were away, then restore or fork the whole machine from that point. Agents can mark points on the timeline and checkpoint the whole machine before risky changes. Other things to try: ssh -t zydra.sh dino # the Chromium dino game via SSH tar -czf - -C ~ .claude .claude.json | ssh -T zydra.sh upload # bring your configs (includes your Claude login) ssh -o ObscureKeystrokeTiming=no -A zydra.sh # lower latency & forward your SSH agent ssh -p 443 zydra.sh # connect from a corporate firewall sshfs zydra.sh: ~/zydra # paid accounts: mount your home Everything runs on one dedicated server with one AMD iGPU, so there are 12 slots. When they're full you wait in a queue and can play Chromium's offline dino game. The paid plan is $40/month (+ tax): 8 GB RAM, 4 vCPUs, 80 GB disk, GPU access, with no time limit and a 3-day money-back guarantee. I'll be around all day to answer questions.
原文Monitor and control Claude Code, Codex, Gemini CLI, Grok, Cursor, and Scoot (local models) from your phone. Permission prompts arrive as a push with the real question and tappable answers. Follow the live conversation feed, drop a note mid-session, or schedule a prompt for 2 AM when your token limit resets. The engine, aaw-core, is now open source: MIT, on GitHub and PyPI. No accounts: pair by QR code, end-to-end encrypted, self-hostable relay. Free for one computer, with every feature.
原文Browser console and diagnostic chamber for Oh-My-Pi (omp) agent sessions — live timeline, steering, real PTY, git/diff/wiki panels, and a panel-plugin platform. Elysia + Preact on Bun. - kabmo-labs...
原文SWARMFORGE is an agentic CI/CD gate for the change path that matters: it reviews a committed or working-tree diff, can group and fix confirmed findings in isolated sandboxes, re-reviews the result, then requires deterministic repository checks before it publishes a PR. It runs with OpenAI Codex/GPT, any OpenRouter model, local open-source models (Ollama, LM Studio), or Claude Code (claude -p, opt-in) — with cost-aware routing measured on an L0–L6 benchmark ladder.
原文View and search Claude Code, Cursor, GitHub Copilot CLI and Codex session logs in your browser. Live refresh, export and local-only processing.
原文阮老师好，自荐一下我做的 Review Inbox。
Hey HN! As an avid user of cmux ( https://cmux.com/ ), I wanted a way to control terminal and agent sessions from my Android phone, so I built Muxer. The app talks to a tiny relay over HTTPS, that then talks to cmux through its local socket, and gives you to ability to: - Answer agent permission requests, questions and plans from a notification - Reply to Claude Code, Codex and Pi chat sessions with text, voice, @ file mentions and photos - Receive all cmux notification, with inline replies, bubbles and background delivery. - Use a live terminal with sticky modifier keys and local scrollback - Share text and links from any app to an agent - Organize workspaces - Manage cmux Cloud VMs and automations - Lock the app with your fingerprint, face or screen lock Enjoy, and let me know what you think!
原文Coding agents rewrite code you never asked them to touch, and the tests still pass. Sonah reads the agent’s Claude Code or Codex session along with the code, then shows on one page: whether each ask was done, what the agent decided on its own, which of its claims its own results contradict, and what is wrong in the code. Every quote is checked word for word. Runs locally through your own Claude login. No account, no telemetry. Free for individuals.
原文Hand off work between coding agents: Claude Code to Codex, Codex to Devin, and more. Every session kept on your machine, searchable and replayable. Open source, local-first.
原文Since our first launch: 22 updates in 17 days, many shipped the same day a user asked. SwarmPane runs Claude Code, Codex, Gemini, Cursor and 9 more AI coding agents side by side, on the plans you already pay for. New: every account's usage limits in your toolbar, each agent in its own worktree, Devin and Hermes, one rules file and MCP setup for every agent, and sessions that survive restarts. Mac, Windows, Linux.
原文MIAO - AgentPal adds animated status-bar avatars to VS Code for Claude Code, KiloCode, Codex, and OpenCode. Track coding progress without switching terminals. Project avatars: Aggregate sessions by project. Five states: Idle, working, waiting, success, error. Customization: Avatars, colors, names, and order. Session details: View status and connection diagnostics. Sound alerts: Notifications for waiting, completion, and errors. Flexible display: Current workspace or all projects.
原文Lore lets a fresh Codex task or Claude Desktop chat find decisions, constraints, and unfinished work in accessible local Codex, Claude Code, and Qwen Code sessions. It returns source references so you can verify the answer. The archive stays on your Windows PC. Free alpha.24; ordinary ChatGPT and Claude cloud chats are outside this release.
原文Agent Standup 是我开发的一个开源工具，把 Claude Code、Codex 和 Mistral Vibe 的会话变成一场实时团队视频会议。 同时开好几个 AI 编程智能体时，很难知道谁在干活、谁卡住了、谁在等你。Agent Standup 读取这些工具本来就写在本地的会话文件，把每个智能体画成一个小角色：主智能体下达指令，子智能体接到任务时加入通话，干活时"共享屏幕"（显示真实的命令输出和 diff），完成后汇报并离开。只有某个智能体需要你时，才会叫你。还可以回放任何一次过去的会话，查看每个智能体的 token 用量。 免费，MIT 协议开源，只读取本地文件，不消耗 token，默认不发送任何数据。支持 VS Code（Cursor、Windsurf 可从 Open VSX 安装）、任意终端和 macOS 桌面应用，界面支持简体中文。 终端试用（Node 18+）：npx @doublegeste/agent-standup@latest --open
I'm an engineer, but I also hate using tools that make things overcomplicated. It's also tough with the new pace of AI-driven development to really get into the flow state, and maintain (still needed!) focus on 1 thing. So I built a tool that I actually want to use every day. Curious if others would find this useful, too, or have any feedback? Like if it's nearly there, what changes would make this a win for your workflow? Details: - It's a Mac app (Apple Silicon) that runs your Claude Code sessions side by side. Each piece of work gets a "context" with its own session, next steps, status, PRs, and links. - The dashboard groups contexts by whose move it is: yours, waiting on a reviewer or teammate, Claude working, or idle. You get a notification when a session needs you or a PR gets review comments. - For non-engineers, there's a Chat view instead of the terminal, a guided Claude Code install, and folder pickers instead of paths. Engineers can stay in the real Claude Code terminal. - It runs the actual claude CLI, and sessions keep running if you quit Perch. You can also import your existing Claude Code sessions. - Everything stays on your Mac. There's no account and no analytics, and crash reports are opt-in.
原文A local menu-bar app for macOS and Windows that shows what every Claude Code session is doing and takes you to the right terminal in one click.
原文Project URL Category Python Project Title LLMSlim: long-term memory for AI coding agents Project An MCP server and Claude Code plugin that gives Claude Code, Cursor and other MCP clients long-term memory on your machine. Messages are stored word for word in local SQLite, and recall returns what answers a question with no LLM call. Highlights One line to install: claude mcp add llmslim -- llmslim mcp. Five tools: remember, recall, listmemories, forget, fitcontext. Messages are stored word for word; nothing is rewritten by a model. MIT licensed, on the official MCP Registry. Example Code markdown pip install "llmslim[mcp]" claude mcp add llmslim -- llmslim mcp or Videos
原文I think I'm not alone with that quota-based AI subscriptions have given me a new hobby: checking how many tokens I have left. How’s my 5-hour and 7-day limits? When’s the next reset? Can this task wait, or should I use a smaller model? I used to check my Garmin for steps and heart rate. Now I’ve built a watch face to track my AI usage limits and reset countdowns too. The idea was to make it very easy to add usage data from other vendors as well.
原文TokenHub brings recorded token usage from Codex, Claude Code, Hermes Agent, VS Code Copilot Chat, and Antigravity into one local dashboard. Explore agents, models, sessions, trends, and estimated costs. Collection requires source approval, and only usage metadata is retained. Free and open source.
原文Stop re-explaining yourself. Goodboy is an ADE that keeps the goal, plan and decisions of every task in one place, so any agent you run already knows what’s done and what’s left. Each task is split into steps, and each step gets the right model and effort. Plug in GitHub, Linear, Jira, Sentry and Slack, review plans and diffs, follow the full log and keep an eye on spend. Works with Claude Code, Codex, Cursor and more, on your subscriptions or API keys. Free and built in public.
原文It sits in your MacBook's notch. Every Claude Code session gets a card, every sub-agent gets a row, and each one has a little pixel programmer. You see who's working, on which model, and who's waiting on you. Free and open source.
原文自荐 DSH A2UI，一个面向 DeepSeek Harness 的独立 A2UI 协议开发预览，供想把交互表单接入 DSH 的开发者试用。 目前使用官方 A2UI renderer 和 schema，支持受限 basic catalog 的严格校验，并提供两种无需账号或 API key 的合成示例： 本地浏览器 流式展示表单、编辑、导出与恢复记录；恢复时不会再次执行历史动作。 可执行 agent-loop 示例：使用真实 DSH AgentLoop、ToolRuntime 和会话渲染组件，通过脚本化本地模型，在 jsdom 中验证“表单操作进入下一轮 agent，并生成预览”的链路。 当前为 0.1.0-alpha.6，针对 DSH 0.2.0-rc.2 验证。示例中的模型、传输和部分宿主结构是合成的，不代表完整 DSH shell、真实模型、完整 A2UI 协议或生产部署已经验收。未提交的表单编辑目前不会跨 DSH 重载自动恢复。项目并非 DeepSeek、Google 或 A2UI 官方项目。
I'm building Multipeer, a desktop app for running agents, tools and working wth teammates on one shared canvas. I often have several agent sessions open at once, but they're all scattered across terminals. It was also hard for my co-founder to see what my agents are doing or jump in to work together. So We built the solution to our own problem. And now we're releasing it to the world. It's a macOS app for now (soon on Windows and Linux). You can invite your teammates over a local network or over the internet. They can follow your view and, if you allow them, they can also take control of your terminal and claude code/codex sessions. The app is still early, and I'm also releasing the iPhone/Android app soon to remotely control everything on your canvas. I'd would like to get feedback from the HN community! If there's any features you'd like to see, please contact me via: shyam@multipeer.sh, we build really fast. I'd love to talk to everyone of you and work on your feedback! Thanks!
原文A small dock that shows how much of each Claude, Codex and Cursor quota you have used, and when it resets. It works with CLIProxyAPI, which switches accounts when one runs out. It also runs in the terminal and in Herdr. Free and open source.
原文Open-source, cross-platform (Linux and Windows) PHP RCE mitigation agent and HTTP WAF. Blocks php2shell session.upload_progress use-after-free exploits (CVE-2026-109618, CVE-2026-109619), session file poisoning, serialized handler mismatch, worker pinning/ASLR bypass, SAPI OOM bailout and PHP-CGI argument injection (CVE-2024-4577).
原文Quai replaces the macOS Dock with one that holds live widgets beside your apps: calendar, weather, music, revenue, and your Claude Code, Codex and Copilot sessions and limits. Native and featherweight. No account, no tracking. $14.99 once, every Mac you own.
原文Built for FDEs, anyone who takes on client projects and manages delivery across many of them, and managers working across many projects in the terminal: one person running many projects with Claude Code and Codex. PMDesk puts every project, agent session, client document and request in one Mac window. See which agent is waiting for you, @-mention client docs, review and send comments back. Your data stays on your Mac.
原文Niyoos sits between your AI agents and LLM providers (OpenAI, Anthropic, Gemini) as a proxy, so it can enforce — not just watch: a tamper-proof audit log, real-time cost tracking per agent, an instant kill switch, and drift alerts when behavior changes. Works with n8n, Zapier, Make, or custom code — point your agent at Niyoos, nothing else changes.
原文LoopCode is an AI coding agent for desktop, terminal, and VS Code. Work on existing projects, inspect commands and diffs, and run checks before reviewing changes. Choose from models available on your plan, with limited free usage to get started.
原文A Claude Code mod that holds subagent fan-out, heavy-context prompts and retry loops before they spend your usage window. It reads your plan's usage, counts agents across every session, and asks in Claude Code's own dialog.
原文A free desktop app for Windows, macOS and Linux. Make an agent for each job, call several into one conversation with @names, and watch their work on a monitoring timeline. Agents run on your Claude or Codex subscription, an API key, a local model or Parallax credits, and ask before anything risky. Reach them from Telegram, Discord, Slack, Signal or email.
原文Paste one line into Claude Code, Codex, Cursor, n8n or any script. Your agent can then put live progress on your lock screen, ping you when it's done, and ask before it deploys, sends or spends: Approve or Deny from the notification. Free, no account.
原文Hatoba is a second dock for macOS. It sits on the left, right or bottom edge of any screen (or in the notch) and stays hidden until you need it. Pin apps, links and 16 live widgets: system stats, music, weather, calendar, notes and more. See at a glance whether your Claude Code, Codex and Cursor sessions are working, waiting or done. Also includes clipboard history, window snapping and with OCR. $14.99 one-time, 14-day free trial, no account.
原文MIT 平台：macOS / Windows（CLI 全平台） 为什么做（背景与痛点） 我同时用 Claude Code、Codex、Kimi Code，还经常通过它们跑 DeepSeek、GLM、Kimi 等国产模型。有一天发现：我根本答不上来「我这个月给 AI 烧了多少钱」——每个工具有自己的用量页，分散、单位不一、没有总账、更没有预算。 它是什么 token-pilot = 你所有 AI 工具的本地账单 + 预算控制器。 一条命令看清今天/本月各工具、各模型烧了多少 token、多少钱；菜单栏常驻 ¥今日 · 预算%；超支系统通知；打转检测（揪出死循环烧钱的会话）；省钱建议（贵模型干轻活迁移、缓存优化空间）。 几个不一样的点 国产模型自动归一：DeepSeek/GLM/Kimi/Qwen 等靠宿主日志 model 字段识别，不用单独配置；定价表对齐 8 家官方目录价（DeepSeek 含高峰/空闲分时） 覆盖 9 个数据源：CLI 日志（Claude Code/Codex/Kimi Code/pi）、SQLite（OpenCode）、zstd 压缩日志（DeepSeek Harness）、trace（WorkBuddy）、云端 API（火山 arkcli）；读不到的工具诚实标注，不假装 数据准确性可验证：自带独立对账脚本（第三方口径重算原始日志逐项比对）+ 39 个单元测试 + 36 项看板算法不变量 零入侵：只读本地日志，不装 hook、不改配置、不过代理，数据不出本机 实测数字（作者本机，2026-10） 本月 810M+ tokens · 12,685 次调用；打转检测抓到单会话 120 次重复重试；缓存命中分析显示本月缓存省了 ¥3,048；压缩潜力量化显示上下文重发还有 $63–95/月可省。 截图 安装：npm install -g token-pilot 或 Releases 下载状态栏应用安装包。
A Claude Code mod that holds runaway agent spend before it happens: it reads your plan's live 5-hour and weekly usage, then asks before subagent fan-out, heavy-context prompts and retry loops - and refuses when nobody can answer. Local and privacy-minimal (no network, no processes, never stores prompt text), fails open, and /agent-guard report shows what it did. MIT, one plugin install command.
原文Halv es el espacio de trabajo de código abierto para Claude Code, Codex, Kimi y GLM. Halv Cloud enruta cada tarea al modelo que necesita: 57.1% menos de costo, con los mismos aciertos.
原文Open-source macOS app for coding agent productivity: agent vs estimated human time, API-equivalent spend, usage limits, and customizable scoring skills.
原文Hi HN, I'm Anish. I built Mnemolith because the AI coding tools I use (Claude Code, Cursor, Codex) either forget my project between sessions, or remember things that quietly stopped being true. Tell the agent "we use Postgres", migrate to MySQL two weeks later, and it keeps building on the old fact. How it works: - It's an MCP server, so Claude Code, Cursor and Codex share one memory per project (keyed by the git repo). - When the agent saves a memory, it can link it to the files it depends on. Mnemolith stores each file's git blob hash. If the file's contents change (committed or not), the memory is set aside and flagged for review, so the agent stops relying on it. - A local dashboard shows what the agent remembers, where each memory came from (tool, session, file versions), and its full history. You can confirm a memory is still true, or retire it and say what's true now. - Contradictions are judged by the agent, not by a similarity threshold. In my calibration, "we use Postgres" vs "we migrated to MySQL" only scored 0.30 similarity, so a threshold can't tell a contradiction from an unrelated fact. - Current memories can sync into a managed section of AGENTS.md / CLAUDE.md, so they get reviewed in PRs. Under the hood it runs on an engine I wrote from scratch in Java: a Redis-compatible server with a built-in memory model and an HNSW vector index. Embeddings run locally (all-MiniLM-L6-v2). Everything stays on your machine. It's early (v0.2.1). There's no team sync yet beyond the AGENTS.md/CLAUDE.md files, and it needs Java 11+ (it downloads a private copy if you don't have one). I'd really like feedback, especially from people running coding agents on real codebases. What does your agent keep getting wrong?
原文Peras puts a line in your Claude Code and Codex week. Set it (peras set weekly 85 pause) and once usage crosses it, the agent's next tool call is refused with a plain reason, so one long task can't burn the rest of your week. Resume whenever you want, from the terminal or the menu bar. Opt-in: your agents can be told their own usage. The CLI and hooks are free and MIT. The native Mac menu bar app is €4.99 once.
原文Hey Everyone, After getting frustrated by Claude rewriting key functions and still making the test pass, was tinkering around with a review tool, that would track the assumptions and decisions the agent took, without me digging through the entire session. So ended up building Sonah. Sonah is basically a code reviewer that not only reads the code and diff but also the agent session that wrote the code, to understand the assumptions, and decisions your agent took along the way. And highlight important parts of the code that need your attention. Sonah uses your existing claude code/codex subscription (claude preferred for higher context window). You can try using below: npm install -g getsonah sonah doctor cd your-project && sonah review --format html --output ~/review.html
原文Termi is a 3D room where your real coding agents (Claude Code, Codex, Gemini, OpenCode and more) work at their own desks. You watch every file read, command and edit live, not a scrolling log. New since our first launch: when an agent hits a usage, rate or context limit, Termi saves its goal, changed files, errors and todos, then hands the work to an agent on another model. A boss agent splits one goal across your team. Deleted agents go to a trash. Now on macOS, Windows and Linux.
原文Meet Glaude — stop babysitting Claude Code. Keep track of your coding sessions, see when Claude needs your attention, and jump back in without constantly switching between terminals. Less context switching, fewer interruptions, and more time building. Claude codes. Glaude keeps you in the loop.
原文Track Claude Code and Codex quotas on Apple Watch. Explore AI-powered Codex reset forecasts on iPhone with Reset Radar. Watch features are free; Pro adds iPhone/iPad widgets and layouts. Claude/Codex usage requires a Mac companion.
原文A free tool that lets AI agents control RDP and VNC desktops and Android phones. Connect your agent to real remote workspaces, watch the same session in a native window or browser, move files, work across multiple displays, and take control whenever you need it. Free to use on Windows, macOS, and Linux.
原文Notchwake turns your Mac's notch into an island. Hover it and it opens: what's playing, your next meeting, a tray you can drop files on, a timer, shortcuts, and your Claude Code sessions, where you can allow or deny a command without leaving what you're doing. No account, no telemetry, nothing sent to a server of mine. Native SwiftUI, light on battery, works on Macs without a notch too. 7 days free, then $12.99 lifetime or $1.99/month.
原文Built this out because I wasn’t happy with conductor, orca, or the other vscode derivative platforms. Please let me know if you guys have any feedback for me
原文The fastest browser for AI agents to run browser automation, built for sharing your logged-in browser state with your AI agents, like Codex or Claude Code, without disturbing you. Zero cost, zero config.
原文Bordet is a shared room where Claude, GPT, Gemini and Grok work on your project together. Each model holds a seat with a role, gives its own answer, then replies to the others where it disagrees. A Chair card shows where they agree and differ, you confirm what is decided, and one memory (a board, a decisions log, your sources) carries the project into every session. Connect Claude Code, Codex or ChatGPT as a seat too. Free to try; pay-as-you-go credits or your own API keys.
原文ccUI 是一个开源的 Claude Code 桌面客户端，给只有命令行的 Claude Code 加上好看、好用的图形界面：一键切换 Claude Code 版本，权限点按钮、改动看 diff，多个会话并排管理。下载安装包就能用，不用敲 npm 命令。支持 Windows、macOS、Linux，界面中英双语。 亮点 一键切换 Claude Code 版本（以防某些版本更新很逆天）：在设置里从下拉列表选最近 10 个版本，或手动输入任意版本号，自动下载并校验，不用碰 npm。想回到最新版也是一键，之后自动保持更新 好看的界面：深色和浅色，Claude 桌面版同款陶土色等配色，毛玻璃、自定义背景图、界面缩放；菜单、对话框、提示都是统一设计 点一下就能操作：权限请求是按钮（允许、总是允许、拒绝并说明原因），Claude 的选择题变成表单，每轮改动的文件在侧边面板看 diff。回答一键复制，消息可以编辑重发、撤销改动或分支成新对话 长对话不迷路：对话右侧的消息导航一点就跳到之前的提问，鼠标悬停显示每条消息的时间 多会话并排：按项目分组，侧边栏标出哪个会话在等你回复，还有桌面通知；需要时一键切到原生 CLI 继续同一段对话 开箱即用：安装包约 108 MB，Claude Code 首次使用时自动下载；开源，MIT 协议 截图或演示视频
我是 PocketCtl 的开发者，向周刊自荐这个项目。 PocketCtl 是一个面向开发者的跨设备工作台，可以在 iPhone 或浏览器上统一查看 Claude Code、Codex、OpenCode 等 AI 编程 Agent 的会话。适合同时使用多种 Agent，或需要在离开电脑后继续跟进任务的开发者。 主要功能： 查看会话输出、工具调用、执行计划和文件改动。 在支持的会话中发送消息、回答问题和处理审批。 管理多台开发主机，并远程浏览会话文档。 Agent 仍在用户的 Mac 或 Linux 主机上运行，使用前需要安装 daemon，主机需保持在线。远程控制能力取决于 Agent 和会话类型；例如 Codex Desktop 会话目前是只读观察。PocketCtl 不提供模型订阅或 API 额度。 数据边界：生产连接使用 TLS；为支持跨设备访问和历史重放，Relay 会处理并存储会话事件，不能视为所有内容都只保留在本机。 项目地址及安装说明：
Enterprise influencer marketing AI agent. You bring the brief and budget; agents source and negotiate with creators in 31 languages, grade every video against your brief and pay creators in 88 currencies. 30% fee, shown upfront. No output, no cost.
原文Running several Claude Code sessions? Bellpip shows each one in your MacBook notch: working, waiting, needs you, done. Pip, a little bronze bell, rings when a session needs you, with a softer chime when a long task is done, and lowers your music for the half second it plays. Click a row to jump to that exact terminal tab in Terminal, VS Code or Cursor. Native Swift, no account. Free for 7 days of use, then $9.99 once.
原文An open-source TypeScript AI agent harness with composable strategies, tools, sandbox adapters, memory, inbox delivery and traces.
原文🚇 Turn distributed traces, GitHub Actions runs, and coding-agent sessions into animated NYC-subway-style maps: GIF, MP4, or an interactive page you can drop in Slack. Fully local, no telemetry. An agent skill for Claude Code, Codex, and Cursor.
原文阮老师好，自荐一个开源项目 「Codex Co²」（ ）: 让多人共享同一个 Codex 会话，一起推进工作。 解决的痛点是，Codex 官方虽然也提供了会话分享功能，但只可以把只读的历史快照发给别人。但如果对方看完后想继续追问、补充信息，或让 Agent 接着排查的话，就满足不了，需要一种能持续交互的协作方式。 Codex Co² 通过生成邀请链接，让一方得以在官方 Codex Desktop 中继续工作，另外几方则可以通过浏览器加入所选会话，查看已有上下文，并向同一个 Agent 发送消息、跟进后续结果。 一个典型场景就是运营与研发的接力排障： 运营先让 Codex 分析用户反馈的日志，得到初步结论，再将这条会话分享给研发。研发加入后，可以直接查看前面的日志分析过程，补充代码背景、追问可疑线索，让 Agent 继续定位问题。运营也能在同一条会话中补充用户信息、跟进排查进展。 主要特点： 共享持续进行的会话：加入后可以继续交互，后续消息与 Agent 回复持续同步。 协作者只需浏览器：打开邀请，经分享者批准后加入，无需安装应用。 由分享者控制范围：只共享选中的会话，可移除成员、结束共享，也可开启逐条消息审批；原生工具审批仍在分享者的 Desktop 中处理。 中文演示视频
Most agent testing either traces failures after customers hit them, or makes you hand-write every test and let an LLM guess if it passed. ClientCoded generates adversarial tests and knows the correct answer, scoring against computed ground truth with SQL, not opinion. It catches failures before production, monitors every live conversation with Slack alerts, and turns real production failures into new regression tests automatically. Connect in minutes with one webhook. First audit free.
原文Limitly: Never get caught off guard by your AI usage limits. Track Claude, Cursor, and Codex usage right from your Mac’s menu bar. Know how much you have left, when your limits reset, and get notified before you run out. 🌱 Fun, interactive themes, 🔔 Usage limit notifications, 🔑 No API keys needed, 🆓 Completely free. Built for people who build with AI. Give it a try!
原文New since the first launch: self-hosted Team features are free with no license key (approvals, SSO, spend-cap holds). Approve risky actions from Slack, Teams, Telegram, Discord, email or the dashboard. Daily money caps and allowed hours per agent, undo for refunds and messages, a one-command install on Linux, macOS and Windows, `agentraas scan` to check an MCP server before you connect it, and keys that can stay in your own Vault or cloud KMS.
原文Run several Claude Code sessions at once, keep notes and prompt history organized per project, search across all of it, and schedule prompts that run even when the app is closed. Get a notification when a prompt finishes, click paths and links right in the terminal, attach your latest screenshot in one click, and readmd files in the built-in viewer.
原文AirHand lets a cloud agent operate Android apps, with a floating view of each step and a Stop button. Open source, no root required. Bring your own Rebyte API key. Android 11+; cloud usage is billed separately.
原文Claude Code forgets everything between sessions, and its defaults are not a careful engineer's habits. developer-os is a free, MIT-licensed plugin that fixes both. Your agent searches an Obsidian vault before work and captures decisions into its inbox while it works; you approve what stays. Rules load in every session, skills cover specs, reviews, audits and PRs, and hooks ask before installs or force pushes. Plain Markdown, no server, no telemetry.
原文Private, tamper-evident notes for your AI sessions, scripts and agents. Write over MCP, REST or one curl; a hook archives Claude Code and Codex sessions. Every note is Bitcoin-anchored, so you can prove it never changed; and much more, please check it out. Free to start.
原文Ask the people you actually trust where to eat, stay, or go. Fluence sends your question to the friends most likely to know, then Trace, your personal AI agent, reads every answer and gives you one clear pick with the why. No fake reviews, no paid influencers, just your people.
原文Unofficial radar for public Codex usage-limit reset signals. Log every event, see the 5-hour vs weekly window split, and plan your heavy runs. The same log goes to Telegram — at most 2 messages a day, /stop any time. Your Codex UI stays the source of truth.
原文Islet turns your Mac's notch into a tool built for developers. It changes with the app you're in: AI commit messages and git status in VS Code, localhost to your phone (or a public link) in Chrome, and `islet run` alerts in Terminal. When Claude Code or Codex needs your approval, a hand pops up in the notch with a chime. You also get PRs with CI status, ports and your Claude usage at a glance. Music, meetings, timer and AirDrop too. ~25 MB, 0% idle CPU, no account, free.
原文A Claude Code plugin that sends hook events (task finished, waiting on your input, a tool you care about was used) to Slack, Discord, email, WhatsApp, desktop voice and web push. Walk away from long runs without missing the moment.
原文目前 agent 的状态工具大多基于 macOS 的刘海，而Agentail 是为 Linux 桌面做的，也支持 SSH 上去的服务器。GNOME 顶栏上有一个小胶囊，显示每个 agent 在做什么。有 agent 等你授权时，胶囊会变成橙色，并显示是哪个项目。点开后每个会话一张卡片，写着在哪台机器、哪个项目、正在做什么。 主要特点： 支持服务器和本地：远程会话通过 ssh -R 隧道传回桌面，在面板里点两下就能添加服务器，服务器上只多一个很小的 hook 脚本。 不影响 agent 工作：hook 发出一条消息就退出。桌面端没在监听时，它立刻返回，不会卡住工具调用，也不会让它失败。 额度一目了然：Claude Code 的 5 小时和每周额度、Codex 的额度，以及各自的重置时间。 本地运行：不用注册账号，没有云服务，也没有遥测。 即开即用：sudo apt install ./agentail0.0.5.deb，重新登录一次，hook、后台服务和面板都会自动配置好。修改 agent 配置文件时只做合并，改之前会先备份，卸载后恢复原样。 支持 Claude Code、Codex CLI 和 Qoder，已在 Ubuntu 24.04 / GNOME 46 上测试（X11 和 Wayland）。非 GNOME 桌面可以用托盘版（agentail ui）。MIT 协议。
Hello HN! Today I am presenting you with Peras, the tool which governs your Claude Code and Codex usage, and enforces limits. I built it because I got tired of constantly having to spam /usage to check my usage, and decide if a task is worth risking the agent running out mid-task. Peras enforces a limit with 2 main tiers. At 75% weekly usage you will get a notification, and at 90% Peras will automatically pause the agent. These limits are the default and they are customizable to your own liking. You may ask "how does it know my usage in Claude Code or Codex?" Well, it's pretty simple. Peras gets the Claude Code usage % from the status line, and Codex's one comes from its local app-server. The "pause" Peras executes is a hook that refuses the agent's tool calls and ends its turn. A new feature I implemented is one that makes agents aware of their own usage. When you send your prompt, Peras attaches the usage %. It sends it at the start of the session, then again at 50, 75, 90 and 100%, or if a reset is coming in 30 minutes, so the agent can hurry up and actually capitalize on the remaining usage, if any. Just to clarify, this feature is informational and the pause is what enforces the limit. Before I switched to native SwiftUI, I used SwiftBar. The problem was that for the mascots patrolling their usage bars, I had to re-encode every menu image 8 times a second! There was also stuff SwiftBar wasn't capable of, like the menu being able to appear over fullscreen apps, pinning it as a floating panel, etc. What is the difference between the free and paid version? Well, it is MIT, first of all. The CLI, the hooks, the SwiftBar menu, and the app's source are all on GitHub, but building the app needs macOS 13+, Node 18+, "xcode-select --install" and then finally "npm run build:app". The €4.99 buys you a signed copy which is notarized by Apple, with updates included. The guard works with no menu at all. To install, just run: npx peras-cli setup npx peras-cli codex setup And now we get to the limits part. For now it's subscription plans only, and when running "claude -p", the pause uses the last interactive reading, so it can come late. Also I feel I have to mention that the guard rail is not a security boundary. It runs inside the agent. The menu is macOS only. The CLI and hooks pass tests on Linux in CI, but I don't use them there day-to-day. Last thing before I end this post, I would really appreciate feedback from you, whether install worked for Codex users, whether you have encountered any bugs, or whatever you feel like sharing.
原文Pick your tools, set your daily agent hours and model class, and see the estimated monthly cost of Cursor, Claude Code, GitHub Copilot, Codex, Devin Desktop (ex Windsurf) and Replit. Prices checked against official pages. Free, no signup.
原文A native macOS menu-bar app for monitoring Codex, Claude Code and Antigravity. Correlate sessions with repositories, inspect lifecycle and provider health, and view usage where supported. Local-first, read-only, free and MIT open source.
原文Connect your own Codex subscription or supported API account to compatible apps through a hosted API. Create separate app keys, choose available models and inspect requests. Provider fees, access and usage limits still apply.
原文Cachejay keeps track of your workday on your Mac (screen text, calls, coding-agent sessions, files and to-dos), files it by project, and gives that context to ChatGPT, Claude, Cursor and Codex.
原文AI coding tools are amazing — until the bill arrives. A single long agent session, a retry loop you did not notice, or a premium model used for a trivial task can quietly burn through your budget. AI Spend Tracker puts every AI tool you pay for into one clear dashboard. Paste or import your usage, set a budget, and see your spend by tool, project and day — before the month surprises you. 100% local-first. No account, no server, no tracking. Your data never leaves your browser.
原文Each Claude Code, OpenCode and omp session becomes a pet that walks along the bottom of Chrome. See which agent needs you, approve its commands and reply from any tab. Runs on your computer: no account, no cloud. $6 one-time.
原文22 focused audits for Swift and SwiftUI apps. Fill in your project once and every prompt updates. Copy an audit into a fresh Claude Code session with a Swift agent skill, then mark it done once you've triaged what it finds. They cover concurrency, security, architecture, performance, accessibility, Liquid Glass, testing, interface copy and more. Most ask for findings by file and line, with a fix for each. Free and open source (MIT). Your project details stay in your browser.
原文换一个关键词或分类试试。
