DEV TRENDS / DAILY DISCOVERIES
Daily discoveries from developer communities, independent makers, and open source.
Open-source tool that aggregates AI model data from multiple sources into a unified, filterable catalog and generates gateway configs (LiteLLM, NewAPI, Bifrost) from selected models. Offers a public model API and a browse page.
Free online image tool centered on color inversion: PNG/PDF/logo inverting, negative-to-positive, black-and-white conversion, plus compression, flip and rotate — 15+ tools. In-browser Canvas processing; images never uploaded or watermarked.
Which Me is a self-exploration assessment site for Chinese and English users, with fun quizzes, daily short tests, and full scales on personality, values, relationships, cognition, and interests. Full scales need server-side payment.
Su Zhe AI sells ChatGPT membership activation via Alipay. Buy a card key, self-activate on the web; delivery usually takes 1–3 minutes, with 1-month subscription warranty. Also offers Gemini, Super Duolingo, and other AI memberships.
AI anime drawing site that generates anime characters, avatars, and Chibi images from text prompts or uploaded photos, with selectable art styles and aspect ratios plus ready-to-use prompts.
Native macOS voice-input tool with BYOK providers (Doubao, OpenAI, Qwen) plus Apple offline recognition on macOS 26+. Multiple customizable text-cleanup modes. No account or server; transcripts and vocabulary stay local. Free, open source.
AI voice dictation desktop app: hold a key to speak, release and tidied text lands at the cursor in nearly any app. Strips filler words and repeats, offers voice translation. Windows, macOS, Linux; free quota resets every 7 days.
Free, no-signup online toolbox covering PDF, image, audio/video and text tools with Chinese/English switching; PDF-to-image and image cropping run locally in the browser, while some tools have usage quotas or file size limits.
A Chinese directory of AI tools and open-source projects, organized by video, image, writing and coding categories with official links, plus GitHub project breakdowns, software download guides and explainers on FFmpeg, APIs and tokens with…
AI video/image creation workbench for text-to-video, image-to-video, image-to-image, with Seedance 2.5 and MiniMax H3. 10 templates turn one photo into a 15s vertical video without prompts. Credits from $9.9; 10 free on signup; 9 languages.
A dependency cache cleaner for developers: scans reclaimable caches left by Xcode, Node, Flutter, Go, Rust, Python and others, flags the risk of each item, and lets you decide what to delete to safely reclaim disk space.
Beginner-friendly GPT/Claude API access service with direct connections from mainland China, no VPN. Register for a key, then follow tutorials to configure Claude Code, Codex and more; monthly plans cost 99/199/299 CNY.
Windows anime desktop pet that tracks AI API balances, quotas and subscription expiry. Auto-checks DeepSeek, Kimi, SiliconFlow and OpenRouter via API key; manual expiry entry for ChatGPT, Cursor.
Private ebook management service for individuals and families, built on Calibre + Vue. Create and share reading lists, sync via WebDAV, convert epub/azw3/pdf, turn epubs into audiobooks or podcasts, and push to Kindle and other devices.
Self-hosted dictionary service with one-click Docker deployment. Imports MDict, StarDict and ECDICT dictionaries, offering token-authenticated lookup and vocabulary-book APIs, a web dictionary UI, light/dark themes and usage statistics.
MyReader is a cross-platform e-book reader for MyBooks. It connects to self-hosted libraries, browses by tags, downloads and reads books, syncs reading data, and supports LaTeX formulas and custom dictionaries. macOS/Windows/Linux/Android.
Turns e-books into a private searchable RAG knowledge base: import EPUB/PDF/TXT, auto-parse, chunk and embed, then answer questions via RAG retrieval. Includes admin for books, tasks and tokens; self-host core-api, worker and web-ui.
Megaphone is an open-source launch toolkit: paste a URL to get a 0–100 launch-readiness score and top fixes, then let AI draft copy for 17 channels like Hacker News, V2EX, and Xiaohongshu, with online checks.
Blue is a Windows desktop tool that organizes files, writes scripts and carries out multi-step tasks within user-authorized scope. The tool itself is free;
Bilingual 3D web tool to explore Hasselblad film back parts. A12/A16/A24 have disassembly and reassembly demos; 13 more models support layered viewing, rotate/zoom/hide and mobile use.
Creative portrait age editor on 1008.ai: upload a licensed adult reference photo, describe an older or younger adult look, then generate in the image workspace. Offers photo tips, prompt examples and an output checklist; credits needed.
A Three.js 3D art gallery in the browser with 18 painting and nature scenes, including Along the River During the Qingming Festival. Features chaptered viewing, free roaming, day/night switching, ambient sound and desktop/mobile controls.
DeepMonkey Studio is a browser-based editor for interactive scene development, offering 2D/3D scene editing, asset management, and SDK/MCP interfaces.
FreeKitHub is a browser-local tool that compresses images to a target KB size and merges, splits or compresses PDFs, plus HEIC-to-JPG and PDF-to-image/Word conversion. Files are never uploaded;
A Finder-like Cloudflare R2/S3 client for browsing and managing bucket files. Offers a PicGo-compatible HTTP image-hosting service for clipboard auto-upload in Obsidian, Typora and similar apps, plus a storage usage dashboard, social-login…
Free open-source Mac voice input: hold Option+Command to speak; on release it transcribes on-device (SenseVoice), uses AI to remove filler words, fix slips and add punctuation, then pastes clean text at the cursor. Includes live captions.
Paid AI video tool turning a single portrait photo into a ~15s Freestyle video: reference-driven body motion, facial expression and lip sync, subject kept recognizable. 480p/720p, downloadable.
koto is a CLI and desktop app that copies, migrates and syncs agent sessions across harnesses: browse sessions, copy any conversation turn into a new session while the original stays unchanged, analyze history and forks in a session tree,…
Upload one clear face photo to generate an about 10-second vertical quince concha / Bby Wow Guy meme short video.
MIT-licensed Mac viewer: open a local folder as one workspace to preview Markdown (diagrams, formulas), PDF, Office, images, JSON, CSV and source, with basic editing, autosave and Git diff/staging. Rust + GPUI Kit, Apple Silicon and Intel.
Turns a single photo (plus an optional outfit shot) into a ~15-second glow-up short video, shifting the look from a home doorway to a nighttime leather street-style shoot.
A Mac utility that restores an app's minimized windows when you switch to it via Cmd+Tab, and requires no system permissions.
Withdrawn by the author.
A free gallery and GitHub of Claude animation examples, pairing previews with authors' published prompts, tools and original sources.
Android bookkeeping app that batch-imports Alipay/WeChat payment and runs on-device OCR plus Qwen3-0.
Self-hosted cloud photo album built with Next.js and shadcn/ui. Offers a waterfall grid with lightbox viewing for an immersive experience, aggregates storage across local, S3, and Telegram, and pairs with the Pixtale Sync desktop tool to a…
Browser-based AI agent for researchers: searches Wanfang and other academic databases, drafts cited literature reviews, and analyzes uploaded experiment data into volcano/PCA plots, result tables and reusable Python scripts.
File Lite is a self-hosted web file manager: a Go backend with a Vue3 frontend embedded in a single binary.
SailGoGo is a free set of 80+ tools that run entirely in the browser, so files are never uploaded to a server and no sign-up is needed.
Open-source Agent Skill + CLI that makes coding agents verify their own UI: axe-core audits in Chromium across default, dark, forced-colors, reduced-motion and 320px states, plus a manual checklist.
MIT-licensed Codex skill that checks code, config and actual commands before writing or reviewing a README, adding setup steps, expected outputs and formatting; based on a study of 200 project READMEs.
upy is a lightweight single-binary web deployment tool written in Go for individual developers and small teams running single-server production environments.
Reviva is a local-first desktop AI agent for learning, research and knowledge work, combining chat, documents, a Wiki knowledge base, notes, review and creation tools, with custom agents, Skills, built-in and MCP tools, and multi-model pro…
Project URL Category JS Project Title Social SDK: one TypeScript client for seven social platforms Project Social SDK is an MIT TypeScript library for adding social features to an app.
CLI that runs WeChat Mini Programs in a regular browser by loading the official WeChat base library and compiler and implementing the native layer itself (message bridge, page stack, native APIs, navbar/dialogs), so rendering, components,…
Rainmeter-style desktop widget engine for macOS, built with Swift/AppKit. Ships Stationery widgets (clock, weather, music, system monitor, notes, to-do), imports .ini/.rmskin skins, plus a Skin Studio live-preview editor. Free, GPL-3.0.
AOCI-CODE keeps a project's AI understanding in a Git-versioned text index: entries record responsibilities, relations, interfaces and constraints of files and DB tables; agents read the index before locating code. Local CLI and MCP server.
Open-source AGPL-3.0 real-time network visualizer: force-directed graph and world map show each connection's country and process. Runs fully local: API on 127.0.0.1, offline geoip, zero telemetry. Flags port scans, beaconing, trackers.
Offline browser-only developer toolbox bundling JSON formatting/validation, regex testing, JWT parse/verify/sign, text diff, chained encoders, and WebSocket debugging. Smart paste, JSON to TypeScript, URL-hash sharing, PWA;
Offline-first frontend workbench turning architecture diagrams into 60fps cinematic camera-move stories: GPU pan/zoom, Sobel edge snapping, bezier pulse flows, synced AI voiceover, <5MB single-file HTML export, ~35KB TS player.
macOS terminal in Rust/gpui that tracks Claude Code and Codex sessions, grouping them by needs-you, error, running, done. Notifies on approvals and reviews each turn's commands, file changes, diffs. Supports SSH, Apache-2.0, no telemetry.
A local-first personal decision system blending Socratic introspection with multi-agent collaboration. Three-layer memory distillation fights context rot; PDR protocol, pre-mortems and anti-sycophancy interrogation counter LLM flattery.
Signal Garden: daily GitHub open-source ranking by 24h/7-day star growth and commits. It manually reviews star spikes and lists only first-time picks, mainly dev tools, AI skills, MCP services and agents. Web, RSS, email; bilingual.
AudioScript captures system audio output, not the mic, and fully offline transcribes it while translating en↔zh into bilingual floating subtitles. No virtual sound card needed, works with headphones; macOS, Windows and Linux.
Open-source Claude Code plugin estimating per-session CO₂e from token usage locally. Shows it in the terminal status bar and a Claude desktop panel beside API cost; per-model and per-token-type factors with cited sources; no telemetry.
Alibaba's open-source multi-task browser agent that works in your logged-in Chrome: extracts data into tables with source links, runs tasks in parallel, saves runs as reusable Skills, and is callable via CLI/MCP.
🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems.
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
No project description yet.
Kronos: A Foundation Model for the Language of Financial Markets
Desktop utility that fetches and processes song lyrics from NetEase Cloud Music and QQ Music, handling batches and exporting lyric files.
SimpleMindMap is a powerful mind-mapping tool.
Unity3D Client And C# Server Framework
🤖 Multi-platform IM AI Agent for Telegram, WhatsApp, Lark, and WeChat. Connects ChatGPT / Claude / Kimi / DeepSeek / Ollama / Pi for auto-replies, community analysis, contact management, and inactive-friend detection.
Using Jev, Lookout watches Slack for time-sensitive messages and takes the action you choose instantly.
A calmer, private Gmail. Emails can look like chat threads, newsletters go in one page, and AI is optional. No tracking, no server, open source.
Keep your agents and builds running on a Mac mini or Studio. Beam brings their Terminal, browser and other app windows onto the Mac in front of you, each as its own window with a Dock icon and ⌘Tab presence. Encrypted between your Macs.
croft pulls data from APIs and files, transforms it with SQL or TypeScript, and keeps it fresh on a schedule. It all lives in one DuckDB file, every table gets a generated type, and Claude Code can drive it. Free, open source, no Python.
Add live widgets, quick actions, apps, folders, search, stats, weather, calendars, reminders, and more into a beautiful second Dock beside your original Mac Dock.
Spin a free 3D globe, play a country’s live radio, or dig to the other side of Earth. Leave a light for a friend, family member, pet or idol you’ve lost. Exploring is free and needs no account.
readywhen is the senior Personal Assistant that works inside WhatsApp. It has its own phone number, its own computer, and a project board that keeps track of everything on your plate. Call it. Message it. Voice-note it.
Meet Wally, your new Scribe agent in Willow. Chat by voice to draft an email/slack/ prompt, rewrite highlighted text and keep refining it in the same conversation.
Upload an MP3 or FLAC and get a complete song folder for YARG and Clone Hero: drums (pro cymbals, 2x kick), guitar, bass and vocals with word-timed lyrics, Easy to Expert, all charted by our own AI model.
A free desktop app for macOS (Apple Silicon), Windows and Linux that puts Claude Code, Codex, Gemini and OpenCode in an MSN-style buddy list. Use separate Git worktrees for separate jobs, or group-chat agents in a shared folder.
Framefields is a high-performance, code-first video engine powered by native WebGPU in Node.js. Built for AI agents and engineers, it replaces bulky creative suites with a single npm package: motion graphics, Photoshop-grade VFX, 3D scenes…
Buying mechanical keyboard switches is a guessing game. Specs and YouTube videos don't tell you how a switch will sound and feel on your own desk.
Lancetta is a native macOS menu-bar app that tells you when and how to use Claude Code and Codex.
Tori AI support agent learns your product in depth from the content you already have, then delivers accurate answers to even the most complex customer questions through a UI you can fully customize.
Own 1 is a fanless personal AI computer that runs two local LLMs and eight agents on your desk, from $3,899 with no subscriptions or per-token costs.
I wanted my humble Claude on mobile to access my WhatsApp messages but the only options I could find either required WhatsApp Business or an actual server to run Baileys on.
A small open model, fine-tuned on a MacBook Air, that writes without em dashes, "not X, but Y" or lists of three, with no system prompt. Answers on 40 held-out prompts, the training rounds, and how to run it.
Hi all, So I have been working on this project KiviDB from February 2026. It started as a project for me to understand rust, having a product with no GC pauses, memory blow up after snapshots and getting OOM on running pods, and that was a…
I'm a software engineer from the east coast (who has lived in Boston, NY, the bay area, UK, France), and I built a task system called 3 & Done. It's the place you trust to store everything you need to do.
STROT (arXiv:2505.01636): compile-time LLM code synthesis with deterministic, self-routing NLU runtimes. PyPI: strot-ai
Opinionated terminal emulator for macOS and Linux: splits, workspaces, file tree and prompt built in. GPU-rendered, written in Rust
Construí una ciudad romana. Un juego de estrategia con población, economía y catástrofes, diseñado para el navegador y el móvil.
No project description yet.
The Codex for research: real papers from OpenAlex, PubMed, arXiv, and Wanfang, across scoping, search, reading, and drafting. Cited reviews, annotated PDFs, and posters.
Vite plugin for UI feedback to coding agents.
TLKRecomp is a static recompilation of The Lion King (SNES), built with snesrecomp
pi-cowboy is an agent orchestrator for Pi( https://pi.dev ) and Herdr( https://github.com/herdrdev/herdr ).
I'm a huge fan of an old project called BulletML which was a way of describing shmup (typically bullet hell) patterns in a kinda weird and clunky XML format.
Live conditions from the weather station on my roof on Long Island, NY.
Each date links to the SEC filing it came from. When a company changes its wording, we keep both versions. No forecasts.
XuanYin Writing 0.1 is a free open-source desktop app for drafting long fiction locally. It logs foreshadowing as entries with buried/recovered status and payoff plans, links each to its exact manuscript sentence and chapter, separates wor…
Parallax is a Jev-powered arXiv literature search and review system that uses set-algebra node graphs to combine multi-route keyword/semantic recall with AI screening and multi-label classification, covering 820K cs.
UnitCtrl mirrors and controls Android phones and tablets from a Mac over local Wi-Fi or USB: live screen, Mac keyboard input, audio, file and app management, screen recording, device stats, plus phone-to-phone control. LAN-only, no cloud.
OhMyGame is an open-source AI game creation tool billed as a one-person AI game studio: you describe an idea, and an agent handles game design, asset generation, development and testing, with Canvas integrated into the game workflow.
The UniXNU team runs iPadOS 17 natively on a Xiaomi Pad 6S Pro powered by Snapdragon 8 Gen 2.
前阵子让 muse 整理云资产，发现早年藏箱子底的项目 ` 和辛苦拿下的 ICP 备案号一直在吃灰,还有 claude 的 tokens,索性拯救了一下尸山代码,搓了一个 macOS 侧边栏笔记：iFolo Note.按我自己使用习惯来的,大家可酌情使用,可提提意见!免费,免费,免费! ### 为什么要造这个轮子？ 每天面对 IDE 和 多 Agent 终端，我在 Mac 上随手记东西一直很割裂： * **重型软件太重**：只是想临时贴段 JSON 、记行命令或闪念，用 Notion 、Obsidian 甚至系统备忘录都得 `Cmd + Tab` 切窗口，打断思路。 * **桌面便签太乱**：贴多了像牛皮癣，过后根本找不到。 我想要的只是一个**平时完全隐形、需要时从屏幕边缘一秒滑出、写完立刻缩回去,AI 时代还支持可控 mcp**的轻量容器。 ### 核心特点 1. **无感边缘呼出**：常驻屏幕侧边，鼠标贴边或快捷键秒开。全屏写代码、开会时随叫随到，不抢焦点，用完即走。 2. **原生丝滑体验**：拒绝套壳卡顿，深度优化了滑出帧率与输入延迟，后台常驻几乎零感。 3. **Markdown 混排直出**：代码片段、Todo 、临时链接所见即所得，无需在源码和预览模式间反复横跳。 4. **不只是临时草稿**：支持双链与标签整理，纯本地 Markdown 存储,目前还没做云端同步、s 5. 3 同步啥的，Mac 侧边栏记下的碎片灵感，随时能在 iOS 端继续整理，数据完全自主。 --- 目前 Macos 端已稳定,后期计划 windows 端，欢迎大家体验,免费、免费!： 👉 **官网**：
OriginalAn open-source life experience community that feeds personal diaries, notes, reflections and others' insights into AI to distill a layered map: domain to scenario to principle to philosophy to life meaning.
Claude Code tool for group treat lotteries on McDonald's China MCP: prices the order, returns a sign-up link, friends join by nickname, winner's address triggers an auto-order. Commit-reveal verifies draws; caps and duplicate-order guards.
Shang'an Xian is a web tool for China's national civil service exam that filters job positions and shows score cutoffs. It removes the 200-result cap, adds fresh-graduate and graduation-status filters, and offers free exam materials.
AI-native TypeScript full-stack scaffold: Vite toolchain, TS monorepo, strict lint (oxlint, ultracite) with AGENTS.md AI rules, an agent skill to pick a stack and generate projects, e2e-tested combos, `vibestart upgrade` template sync.
Try another keyword or category.
22 repositories were already featured in the last 3 reports, so they are not repeated in today’s discoveries.
The job search that runs on your machine. AI job application framework built on Claude Code: evaluate postings, tailor CVs, write cover letters, prep interviews. Fork it and own it.
Orca is the place for working with a fleet of agents (e.g. Grok Bot, Claude Code). Run any coding agent with your own subscription. Available on desktop, mobile and remote runtime.
The open source coding agent.
DEV TRENDS / COMMUNITY
Discuss this edition
Sign in with GitHub to share a thought, question, or useful follow-up.