DEV TRENDS / CATEGORY
109 new projects first discovered during Sep 6, 2026 – Sep 12, 2026, deduplicated across 8 sources.
Native Mac tool: type @@ in any text field to send screenshots, selected text, clipboard items or Finder files to your own Claude Code/Codex agent, preview the answer, then Tab it back in. Also screenshots, recording, clipboard history.
Mobile Easy Use lets AI agents access Android/iOS app runtimes via MCP for debugging and validation. It adds UI automation, business/config reads, method hooks, call stacks, timing/memory, value/field swapping. Based on Frida; MIT.
Deck: MIT-licensed macOS kanban app managing multiple terminal CLI agents like Claude Code and Codex. Cards map to real tmux sessions, grouped by project, with cross-project alerts for sessions awaiting input.
Go CLI managing Linux terminal proxy env vars: prox on validates the proxy before setting HTTP(S)_PROXY/ALL_PROXY for the shell and children; off restores prior values; status/check report state and connectivity; run wraps one command.
Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows
UI design workflow for AI coding agents: turns spoken requests into plans, gathers licensed assets, enforces a design contract (semantic tokens, Do/Don't rules, motion), and does browser-based desktop/mobile/keyboard/motion checks.
Open-source architecture visualization for AI coding changes: links agent-declared activity to system modules, file ownership, source evidence and verification results, and renders standalone HTML views needing no server.
📄 Awesome OCR multiple programing languages toolkits based on ONNX Runtime, OpenVINO, MNN, PaddlePaddle, TensorRT and PyTorch.
claude-red is a curated library of offensive security skills designed for the Claude skills system. Each skill is a structured SKILL.
RuwoScan: AI-led automated penetration testing via natural language. Multi-agent architecture autonomously runs reconnaissance, request sending, vulnerability validation and evidence collection, deciding targets, methods and when to stop.
NanoAvatar runs lifelike digital humans on older phones without cloud GPUs, with fast, smooth generation. Inference code and weights are released, plus a video and Android APK; record your voice to drive the avatar offline, no API key.
Mementolink creates public memorials by writing text and a photo permanently onto Bitcoin, not a link. No wallet or bitcoin needed. Text from $19, formatted text $39, photo $109. Once published, on-chain text cannot be changed or deleted.
macOS menu bar tool mapping right Command/Option gestures and Mac body taps to actions like launching apps, custom shortcuts, screenshots, volume, DND and window controls. Offline local configs, Chinese/English UI, macOS 14+.
Cortex is an open source API knowledge layer that turns OpenAPI, AsyncAPI, GraphQL, gRPC, and OpenRPC specs into interactive documentation, typed SDKs in 11 languages, and MCP servers for AI agents.
QApilot MCP automates tests on real Android devices and emulators by talking to Claude, Cursor or Codex. Describe a test flow in plain English, no Appium code required. Your agent lays out the steps. QApilot runs the steps.
Stop writing a parser for every service. Pascal’s Pager uses AI to turn raw webhook JSON into clear iPhone Alerts. Give each service a private URL, tell Pascal what matters, and get a summary with key facts and useful links.
Stackness is where developers and vibe coders show the tools, workflows, and approaches they build with. Build a stack profile you share as one link, post your moves, follow people whose taste you trust, and watch tools rise and fall.
Client-side autofill for Ant Design job forms: Playwright extracts the schema, any LLM returns a JSON plan, a local script fills fields. No API calls, fees or uploads; matches label text, retries on React re-renders, manual submit.
💫 Toolkit to help you get started with Spec-Driven Development
Open-source vocabulary site for Chinese speakers abroad: built-in or imported word lists, daily FSRS spaced repetition with known/vague grading, and AI pre-filled entries covering senses, examples, collocations and mnemonics.
CLI tool that reads and writes local Jianying/CapCut draft files to batch add subtitles, style captions, edit timing, speed, volume, templates and animations, and split long videos into shorts. JSON in/out lets any LLM agent drive it.
Project URL Category JS Project Title Official MCP server for the Taskade workspace Project MIT-licensed MCP server that connects Claude, Cursor, VS Code, or any MCP client to a Taskade workspace.
Single-file static dashboard aggregating One Person Company intelligence in nine categories, with full-text search, filters, top-10 ranking and card/timeline views. Zero dependencies, offline-capable; MIT-licensed data.json.
ContextOS is a macOS app and MCP server for Codex: it stores a Git-tracked project graph, serves path::symbol locations instead of full files, compresses terminal logs, and gates completion on test results, cutting long-task context ~60%.
codereview-ai: self-hosted AI code review platform. It posts line-level PR/MR comments via LLM, Semgrep, and agentic reasoning. Webhooks trigger reviews, backfill missed PRs, notify DingTalk/Feishu/WeCom. Docker deployment, Vue 3 admin.
Inbucket is an email-testing service that can be embedded in your app to capture and inspect outgoing emails, with a built-in web interface.
Converts PDF, DOCX and HTML to Markdown in-browser; text-layer PDFs never upload. Edit/copy results, download files or batch ZIP. Cloud OCR for scans; Markdown-to-PDF, REST API, MCP for automation. Local use is free.
A programming game where you implement the core gameplay mechanics of various games yourself while playing through a game developer's story partly drawn from real experience.
Devin Voice lets you talk to Devin, Cognition's AI software engineer, and have it ship the work. Speak a task out loud and Devin plans, codes, and delivers.
Jackalope brings Codex, Claude Code, Grok, and OpenCode into one shared workspace. Run tasks in parallel Git worktrees, keep project context together, and review changes before applying them.
The gap between 1.5x and 100x isn't speed, it's trust!! AI agents generate code 100x faster than humans write it. But your team can't review it all.
Cadenya is not a framework you bolt into your application stack. It's a hosted agentic loop. You connect tools using specs you already know, like OpenAPI and MCP, and Cadenya runs the agent for you.
Kernel-level anti-detection browser that alters fingerprints in C++ instead of injecting scripts. Exposes a standard Playwright BrowserContext over CDP, so existing Playwright scripts run unchanged. Free tier: unlimited local profiles.
OpenAI-compatible reverse proxy that gives any LLM real-time web search with source citations. Residential proxy routing plus TLS fingerprint spoofing bypass Cloudflare/WAF blocks; FastAPI async, pip and Docker install.
Never stop coding. Free MIT AI gateway: one endpoint, 352 providers (150+ free), 1200+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot.
An accessible, humorous Java learning guide covering Java fundamentals, concurrency, the JVM, enterprise development, and interview preparation, aimed at self-taught learners and job seekers.
JeecgBoot is an enterprise AI low-code platform that generates full-stack code and whole systems from prompts. AI Skills create workflows, forms, reports, and dashboards; built-in AI offers chat, knowledge base, orchestration, MCP plugins.
ReadCLI is a terminal novel reader supporting TXT and EPUB, with a local bookshelf, reading progress saving, bookmarks, search, customizable features, and several IDE-style interfaces.
MERVYX is a marketplace for verifiable outcomes from AI agents, APIs and automation services. Providers list services with pricing and acceptance criteria; buyers post tasks with budgets for quotes. Currently 0% commission.
VibeToken is a unified API platform for devs and AI teams to access Chinese LLMs via official channels. It offers domestic direct connection, pay-as-you-go, centralized model selection, API key/usage management, chat/image/video generation
Git-backed docs with real access control, in one markdown with folder-level access where teammates, outsiders and AI agents each see exactly their folders.
Most software is designed twice: a picture first, then again in code. Intent drifts in between. Modeinspect is the AI design canvas for working on the real thing.
Desert Ant Labs builds small AI models that run on your phone or browser, no internet, no per-use cost. Instead of one big model doing everything, they make small ones, each nailing one task, across speech, text, and vision.
Nova is the AI Operator that builds and improves your AI customer experience team, no developer required. Describe in plain language how customers should be handled and connect your tools.
49Agents IDE is a 2D canvas where every agent, terminal, and machine you own lives on a single map that you build yourself.
Ajiez is a free online toolbox with 70 tools in 8 categories. All processing runs in the browser: AI background removal (RMBG1.4) and speech-to-text (Whisper) use downloaded models. No sign-up, no limits, no watermarks, works offline.
Free cross-platform C++ photo browser with cascading grid, GPU rendering, hardware video decoding, timeline browsing, library folders, filename search, favorites, EXIF/GPS with offline city matching, encrypted nearby sharing.
AlphaGenome Atlas is Google DeepMind's AI-powered map of how genetic mutations may affect human biology.
Free dead pixel test tool that uses full-screen solid colors to check monitors, laptops and other screens for dead, bright, dark pixels and display issues.
rpi is a Rust-native implementation of the Pi agent, following a library-first design that splits providers, agent loop, tools, sessions, and terminal UI into separate crates.
CodexBar is a macOS menu bar app for Codex. It shows quota, reset time, token usage; auto-uses before reset; tracks tasks; notifies on completion/approval; prevents sleep during tasks. Swift/SwiftUI, macOS 15+, GPL3.0. bobzebedy/CodexBar
Cross-platform native email client in Rust and Slint, supporting IMAP, SMTP, JMAP, CalDAV calendars and contacts. No WebView; HTML mail bodies render via Blitz in Rust. Minimum ~20 MB RAM. Linux, macOS, Windows, early Android. AGPLv3.
wenbiao: 5 copyable prompts—tender parsing, score-point breakdown, outline, chapter drafts, bid-rejection checks. Missing facts marked 'to be filled'; no fabricated credentials; human review required. Open prompts/workflows, not SaaS.
Open-source multi-agent dev env for Codex, Claude Code, and OpenCode. Terminal/chat views, Git Worktree isolation, kanban, Tailscale phone access. CLI for agent task/worktree/session control. Codex imagegen. Windows/macOS/Linux; AGPL-3.
Open-source macOS workbench for many Claude Code sessions: session cards, a fork DAG for alternatives, a project kanban. Sessions read each other's screens and delegate tasks; forks use Git worktrees; notifications fire only on decisions.
Codex Memories: a preview local memory tool for Codex. Stores project rules and decisions in Git with a local index. An API-key-free writes fake memories, recalls them in a new process, and cleans up, so you can try before installing.
DSHX: an open-source TypeScript/React toolchain for DeepSeek Harness plugins. Vite-based, with scaffolding, diagnostics, typed Host–Client APIs, and a community plugin/compatibility site. MIT, Preview; needs compatible DeepSeek Harness.
Local-first AI coding agent desktop: Electron + Rust host core + pi Agent Harness + user-installable plugins
PhoneCam turns Android 12+ phones or iPhones into Mac cameras via USB/ADB, Continuity Camera, or UVC. Features: preview, rotation/mirror/crop, CLI, Syphon output to OBS. Video stays local; no root/APK. MIT; macOS 13+ source-build preview
Runner is a local AI agent orchestrator for multiple Claude Code / Codex in one native desktop app. It uses runner configs, lead-led crews, missions, PTYs, local append-only logs, askhuman, cross-vendor MCP server, Rust/GPUI, GPL3.0.
YYLO is a CLI that governs AI coding agents: each task freezes the protected branch SHA, runs in its own git worktree, and must clear validation, merge and release-readiness gates. Review is risk-tiered; runs stay auditable via receipts.
Harden AIF is a free, local security tool for AI coding agents. Its post-trained model checks tool calls before they run, using your request and session context.
From the team behind Gatsby, Mastra is a framework for building AI-powered apps and agents with workflows, memory, streaming, evals, tracing, and Studio, an interactive UI for dev and testing. Start building: npm create mastra@latest
Global high-purity dynamic residential proxy network for overseas expansion, cross-border D2C e-commerce and AI deep search; offers anti-blocking infrastructure, 1GB free developer trial, and a shield-bypass benchmark dashboard.
Minimal macOS window manager for multitasking and AI coding: hold Ctrl, point at a zone, release, and windows snap into place—no shortcuts needed. 20+ layouts with draggable, remembered ratios. 2–4 splits free; Pro ($5) unlocks 5+ splits.
开发者自荐浏览器工具“AI开源项目雷达”，提供日/周/月趋势榜、爆发、新项目等入口，展示用途/语言/适用人群，建议回上游核对。收录开源AI项目，排名非审计。
Original无需注册、免费即用的 AI 一键抠图工具，秒级输出边缘干净的透明 PNG，支持电商人像/发丝细节与 4K 高清，提供批量处理、免费试用与开发者 API
OriginalThe Freight Class API for shipping & logistics software Turn any product into its NMFC code and freight class in one API call - built for TMS, freight quoting, and AI logistics apps. Plus free carrier terminal locations for 10 major carriers and 4,400+ global seaports, all under one key.
Codex task status lights for the NuPhy Air65 V3 on macOS
免费开源 CSV 中文乱码修复：浏览器本地转换 GBK/UTF-8 BOM，可离线使用。Privacy-first offline CSV encoding converter for Excel.
OriginalFree, open-source YouTube transcript downloader: a Chrome extension that batch-captures videos, playlists, and channels into timestamped Markdown you own — plus a public transcript archive at transcriptly.libmap.cn.
Curated GPT-6 Astra demos: games, Blender & 3D, websites, and video creation. Original posts, creator credits, and live projects.
Interactive typo fixer for npm, pnpm, and yarn commands
Open-source agent runtime for persistent cognition, structured context, and concurrent work.
CLI for browser automation: let any AI agent drive your real Chrome — logged-in tabs, zero deps, no relaunch, no MCP server. A11y snapshots + element refs, CDP screenshots, network capture. 让任何 AI 智能体(Claude/Kimi/Qwen/DeepSeek/GLM)驱动你真实的 Chrome 浏览器——已登录会话、零依赖、无需重启。
Original<img width="5120" height="2880" alt="Image" src=" />
Native macOS archive utility built with SwiftUI. Local-first, universal Apple Silicon / Intel support, English / 中文 / 日本語.
OriginalAI Context Linker is a privacy-first context compiler and open schema. It lets AI assistants understand local projects through minimal, auditable context bundles without uploading the codebase. Agent skills are optional interfaces over its standalone CLI.
Caravel — a local-first workspace for AI coding agents. Delegate tasks, follow progress, and review code across projects.
Digital asset flows you can build and deploy with AI.
Spatial canvas IDE for AI coding agents
AI Sales rep for your website visitors
The Cloud for AI Agents
Automated Penetration Testing for AI-Built Apps
An app that keeps Codex on the job.
The world's first open-source AI-native vector design tool and the first to feature concurrent Agent Teams. Design-as-Code. Turn prompts into UI directly on the live canvas. A modern alternative to Pencil.
GradeClarity 免费网页工具，适用于百分制、期末按比例计入总评的课程。输入当前成绩和期末占比，可查看目标总评所需期末分、可能总评范围及可达目标。
Original将操作原子化并定义DSL，支持自由组合图片/视频/音频处理步骤，生成可分享链接供他人直接使用；灵感源于境外工具慢且需多步操作，故自建工作流工具。
Original自制日语练字工具「漢字ドリル」（kanjirenshu.com）：选假名/汉字/姓名生成A4字帖PDF，支持描红、笔顺、课堂材料，免费免注册，浏览器本地处理。
Original自荐文章探讨AI代码迁移与旧系统知识保留。从Joel Spolsky的警告出发，结合Bun、Wiz迁移案例及自身JS→TS→React经历，讨论AI降低重写成本后仍不可丢弃的东西，重点论述旧实现的验…
OriginalFocuh 是 Mac 免费 ADHD 待办/专注工具，v0.7.0 已签名公证，支持 Apple Silicon+Intel。
OriginalRun multiple local coding agents on your machine
See code through a new lens
The open source coding agent.
iOS 17 重构设备通信架构，导致 Xcode 15 无法下载开发者镜像，libimobiledevice 工具在 iOS 17 设备上无法获取性能数据，开发者与测试工程师均受影响。
OriginalCatLitter猫砂：可逆冗余代码清理工具，扫描隔离后观察再永久删除，支持字节级还原。自动适配语言框架，保守豁免降误报，附缓存清理janitor。Python/MIT，37测试通过。
OriginalC 划词翻译工具：MATLAB 等取词困难场景也可用的热键翻译 这是一个面向开发者的 Windows 划词翻译工具：选中英文按热键弹出中文译文。剪贴板 + UIA 双通道取词，MATLAB（Java 控件）等传统取词方案失效的场景也能用；终端类前台自动改用 Ctrl+Insert，避免 Ctrl+C 误杀运行中的任务。支持 DeepL、百度与任意 OpenAI 兼容大模型，翻译时自动保护代码、路径与 URL；附截图 OCR 与输入翻译，绿色免安装，MIT 开源。 亮点 MATLAB（Java 控件）等取词困难场景可用：剪贴板 + UIA TextPattern 双通道取词，UIA 走独立子进程防挂死 终端防护：终端类前台绝不注入 Ctrl+C（SIGINT），全部改用 Ctrl+Insert LLM 翻译代码保护：代码块/路径/URL/标识符先占位再还原，恢复失败不静默 截图 OCR 翻译：Windows 内置引擎，免费离线，深色截图自动反色增强 多行输入翻译、结果窗复用、失败自动降级与 5 分钟缓存省额度 绿色便携（zip 约 870KB）、密钥 DPAPI 加密、WebView2 系统自带运行时 项目截图/演示 README 顶部有 5 秒演示 GIF（选中 → 热键 → 译文 → 复制）：
OriginalTranslate for Developers：开源 Windows 划词翻译，通过剪贴板+UIA解决 MATLAB 选区问题，终端前台自动改用 Ctrl+Insert 防误杀进程；
Original🌩「自选优选 IP」测试 Cloudflare CDN 延迟和速度，获取最快 IP ！当然也支持其他 CDN / 多个解析 IP 的网站 ~
Original你好，我是 RunWhale 的开发者，想自荐一个基于 DeepSeek Harness 的开源手机编程工具。 RunWhale 让你在只有手机的时候，也能把想法做成可以试用的应用：描述需求，让 AI 帮忙修改代码，再直接在同一台手机上预览结果，无需电脑参与编辑和预览。 可以新建项目或导入已有仓库，查看和修改文件、检查 Git 变更，并继续之前的对话。目前主要面向 Web 和 Expo / React Native 项目。 它的一个特点是，项目文件、编程助手的工具和预览环境运行在手机上，模型推理使用用户配置的远程服务。它不是远程控制电脑，也不是完全离线运行的 AI。 项目基于 Expo、React Native 和 DeepSeek Harness，原创代码采用 Apache2.0 协议开源。 演示视频： 欢迎试用和交流，感谢考虑收录！
Originalctty is a terminal application for managing SSH, serial, Telnet, and similar connections through a terminal-based graphical interface.
快说输入法：Windows离线语音输入，本地识别零上传。快捷键说话出字，支持会议逐字稿、声纹过滤、悬浮条和Webhook推送。Windows版已上线，macOS开发中。
OriginalSeatingKit 是婚礼宴会座位安排工具，提供可视化工作区，支持导入宾客、摆放桌位、分配座位、查看建议或交换座位，并可导出 PDF/PNG。免费版最多100位宾客，支持多设备，数据本地保存。
OriginalJR 自荐免注册中文 SQL 试验场：基于 DuckDB/WebAssembly 在浏览器运行，提供合成事件数据与可修改查询。默认示例计算请求数、5xx 错误率，可练习区分流量与失败比例；
Original快咔截图是Windows免费开源截图工具，纯本地运行，无账号、云端、埋点。支持0.24秒冻结、像素级框选、八种标注、本地OCR、钉图置顶、任意方向长截图及跨屏混合DPI。
OriginalA powerful AI agent with fewer restrictions
Try another keyword or category.
