项目大厅

探索全部独立开发者项目与共创求助

全部类别 AI 应用 AI Agent 开发工具效率设计创意游戏工具

已上线

开发工具claude-code

BagIdea Office

Your desktop wallpaper becomes a living AI-agent office — real Claude Code sessions that walk, work, delegate and hold meetings.

BagIdea Office turns your desktop wallpaper into a living HD-2D office where every pixel-art character is a real Claude Code session. You play CEO: give an order, a Director agent breaks it down and delegates, and you watch the actual work happen — characters walk to their desks, open files, and run tools in real project folders. A glowing monitor means a live tool call is running; a character walking over to Security means a real permission request. Nothing on screen is canned animation. The office extends itself: agents propose plugins and skills (you approve each one), accumulate shared knowledge in OFFICE.md, and even hold meetings on their own that can turn into written project proposals. It talks over voice, Telegram, Discord and LINE, speaks 14 languages, and lets each agent run a different model brain — Claude, GLM, DeepSeek, Qwen, OpenAI, Gemini, or local Ollama. The project is heavily maintained (241 stars, 98 releases) and MIT licensed. The author's own disclosure: the app was itself built with Claude Code. curated by the VibeFix editors

0

0

15

0

已上线

开发工具

termist

Mission control for your coding agents: Claude Code, Codex and OpenCode side by side in one terminal.

termist is a terminal mission control for people who run multiple coding agents at once. Every session becomes a card with a dot: running, waiting for you (a permission or a question), done-unseen, done-seen. Jump straight to the agent that needs an answer, send follow-ups without opening its terminal, resume sessions by name, manage projects as tabs — and the agents keep running when you close the window, owned by a small daemon. The author's disclosure: AI coding tools were used throughout the project — implementation decisions, testing and project direction are the author's. It is at v0.1.4 and moving fast: GitHub PR review (comments, line threads, reviews) already landed, installable on macOS, Linux and Windows. Plus the best-looking terminal UI of the year: Istanbul scenes (Galata Tower at dusk, the Bosphorus) that change with the hour. curated by the VibeFix editors

0

0

12

0

已上线

娱乐

Cryptid Field Station

A vibe-coded cryptid-reporting station: file a sighting, an AI editor scores it, a ranger approves it.

Cryptid Field Station turns hiker sightings into a verified field guide — a Sanity Challenge submission that is genuinely fun to use. File a sighting with story, location and evidence; an AI credibility check plus a real weather cross-check (Open-Meteo pulls the actual sky that night) score the report; then a human ranger approves or rejects it through a real-time triage board — with every transition audit-logged. The workflow even rejects Bigfoot blurs. The author built it entirely by vibe-coding on Sanity: the field guide, a Ranger Dispatch board with live subscriptions, a custom Sanity Studio input that turns trail-cam video URLs into 6-frame GIF strips, and the review process modelled as data (stages, guards, effects). Seven cryptids tracked, twelve sightings filed, eight verified and published. curated by the VibeFix editors

0

0

14

0

已上线

娱乐claude-code

Vibe Games Arcade

4 款一句话 prompt、单次生成的完整游戏：LOL 峡谷仿真、MC 体素沙盒、PVZ 与杀戮尖塔。

Vibe Games Arcade 是一个 one-shot vibe coding 实验：4 款游戏全部用 Claude Opus 5.5、单次一句话 prompt 直接生成，零后端依赖，纯前端点开即玩。英雄联盟峡谷仿真（精准防御塔仇恨、三路兵线、A* 寻路、完整技能冷却）、我的世界 3D 沙盒（柏林噪声无限地形、体素物理、第一人称漫游）、植物大战僵尸（植物图鉴、阳光经济、波次攻防）、杀戮尖塔肉鸽卡牌（Act 地图、手牌循环、遗物系统、怪物意图预警）。作者在仓库 README 里明确写道：「本仓库收录的所有游戏作品均为使用 Claude Opus 5.5 进行一句话 Prompt 单次生成 (One-Shot Vibe Coding) 的实验成果。」游戏大厅直接展示了每款游戏的原始 prompt——最短的只是一句话「尽可能利用你的现有能力复刻 Minecraft」。单次生成做到「能玩的游戏」仍是极限挑战，这个合集就是一份公开的、可亲自验证的能力边界报告。 curated by the VibeFix editors

0

0

17

0

已上线

开发工具claude-code

MechFaber

Describe a machine in words — AI picks the parts, writes the CAD and firmware, then simulates it all together.

MechFaber is an AI machine-design engine: you describe the machine you want in plain language, and it searches vendor catalogues for motors, bearings and controllers you can actually order today, draws the CAD around the suppliers' own 3D models, validates fit and load, lays out the full electronics loom, writes the firmware, and then runs the entire machine in simulation — real firmware binary on an emulated MCU (Renode) coupled tick-by-tick with physics (MuJoCo), sharing one battery model. When something is wrong, you reply in a sentence and the whole loop re-runs. The author's own description of the build: Claude Code did all the design work including the firmware. Built over roughly 3 months with a split-agent setup — separate agents for research, calculation, CAD, electronics, firmware and simulation, after a single do-everything agent failed. Clone machines from the library and rebuild them locally, or publish your own. Free while it is early. curated by the VibeFix editors

0

0

14

0

已上线

社交社区replit

Faded

A photo social app where every photo expires — images fade and blur as their shelf life runs out.

Faded plays with digital scarcity. Every photo has a shelf life — by time or by view count. The closer it gets to expiring, the more the image fades and blurs; when it expires, the photo disappears and only the comments remain. The author's own confession on Show HN: 'And, yes its vibe coded, built on replit.' The live site already has real users posting, and the fading-photo mechanic is genuinely playable. 'Nothing lasts forever' as a product mechanic — a small, honest idea executed live. curated by the VibeFix editors

0

0

12

0

已上线

娱乐v0-by-vercel

144 Archetypes

A personality assessment fusing 16 cognitive types × 9 Enneagram types into 144 profiles — written entirely by v0.

144 Archetypes crosses the 16 MBTI cognitive types with the 9 Enneagram types to produce 144 deep personality profiles — four-force models, stress responses, energy patterns — generated from a 66-statement assessment. Mai, a Tokyo radio personality and product manager with no engineering background, built the entire app using v0 by Vercel. 'I wrote the prompts in natural language, and v0 wrote the code.' It's the textbook non-engineer-solos-a-complete-product vibe coding story, with a 16-type browser and Enneagram grid that hold up on inspection. The site has been live for roughly ten months and is still maintained and content-complete. curated by the VibeFix editors

0

0

9

0

已上线

开发工具boltnew

GHOSTVAULT

Enterprise-grade dark-web breach monitoring with zero-knowledge architecture — built in 72 hours on Bolt.new.

GHOSTVAULT is a breach-monitoring platform that aggregates 5+ leak data sources (HaveIBeenPwned, DeHashed, IntelligenceX and more), encrypts everything browser-side with AES-256-GCM so even the operator can't see your data, and fires real-time dark-web alerts. The author, Alex Sofroniev, built it to prove 'enterprise-grade cybersecurity could be built rapidly using modern no-code platforms' — 72 hours, $0 development cost, and the project grew so technically complex it started hitting Bolt.new's own platform limits. Pricing runs from €6.90/month Pro to $299/month Enterprise, and it entered through Bolt's World's Largest Hackathon. curated by the VibeFix editors

0

0

12

0

已上线

生活方式lovable

Plinq

Women's safety background checks — type a name and number, get a criminal-record risk score in seconds.

Plinq is a women's safety app from Brazil with a simple, urgent idea: type a full name and phone number, and it checks public criminal and judicial records, returning a red/yellow/green risk score in seconds — plus a one-tap emergency alert and a mutual-aid community. The founder story is the point. Sabrine Matos, a growth marketer with no engineering degree, built the whole thing on Lovable — website, desktop app, scoring logic, public-records integrations, backend workflows — in 45 days. 'If Lovable didn't exist, Plinq would never have seen the light of day.' Three months in: 10k+ users, roughly $456K in annualized revenue, 300% month-over-month growth, and a claim of 200+ dangerous situations avoided — with coverage from G1 Globo, Exame and Marie Claire, and a pre-seed round in progress. curated by the VibeFix editors

0

0

9

0

已上线

游戏claude

Linnaeus

A daily taxonomy speed-run game: classify a species down the tree of life, 10 seconds per choice.

Linnaeus gives you one species a day and a stopwatch. Classify it from kingdom down to species — 10 seconds per choice, faster answers worth more — and climb the daily and monthly leaderboards. The author's pitch is endearingly plain: 'A friend gifted me a week of Claude, and this is what I got in 3 afternoons of full vibe-coding.' Three afternoons, and out came a polished daily puzzle with a streak system and a whole tree of life to race through. Stack is Svelte 5 on Cloudflare Workers with D1 for data — a tidy, serverless setup for a game that only needs to wake up once a day. curated by the VibeFix editors

0

0

9

0

已上线

开发工具claude-code

jpm

A fast, small, secure-by-default JavaScript package manager written in Rust — every line coded by Claude Code.

jpm is a JavaScript package manager written in Rust, and it makes a bold claim: up to 15x faster than npm, a ~2 MB binary, and security by default. Install scripts need explicit approval, and fresh versions of a package get a one-day cooling-off period before you can pull them. The story behind it is the real hook. The author directed Claude Code, running Claude Opus 5.5, to write every line — its own TLS and crypto included — then spent five days with the agent installing popular packages and hardening the implementation against real-world breakage. The site is brand new (the repo landed 2026-09-28), and the homepage leads with the honesty the genre needs: 'Every line written by Claude Code. Directed by one engineer. Checked against everyone else's tests.' curated by the VibeFix editors

0

0

9

0

已上线

开发工具claude-code

Incident War Room

Real-time incident response for on-call teams: a live timeline, a SEV1 gate needing two on-call leads' approval server-side, and an AI-drafted postmortem.

Incident War Room is what you wish you had open at 2 AM when something breaks: everything about the incident lives on one page — a live timeline every responder sees update in real time (Sanity's `listen()`, no polling), a status page entry that publishes itself once an escalation is approved, runbooks for the incident's severity right on the page, and an AI-drafted postmortem when you mark it resolved. The centerpiece is the SEV1 escalation gate, and the author designed it specifically so a judge couldn't fake it. Raising an incident to SEV1 needs sign-off from two different on-call leads, enforced server-side in an `approveEscalation` Server Action — the approver is whoever the signed session says is logged in, must be an on-call lead, and can't approve twice. A Sanity Workflow tracks `escalationApproval` as `pending → approved / rejected` on a Kanban board in Studio. The build diary is the best part: the first version of the gate took the approver id from a hidden form field, so Alice could "Approve as" herself and Dana and push a SEV1 through alone — Claude Code caught that itself during a status check and it got fixed. Built for the DEV × Sanity challenge "Path Two: Vibe-Code Something Strange" with Claude Code, from an empty directory to a working demo — and the author is honest about where the model got stuck: Next.js 16 was newer than the model, so the repo's `AGENTS.md` points the agent at the docs bundled in `node_modules/next`; the `listen()` API ignored GROQ projections and two clocks (UTC vs local) disagreed, both found by driving the live site with Playwright. Resolving an incident freezes the timeline and sends it to Groq (`openai/gpt-oss-120b`) for a root cause and action items — told to say "unknown" rather than guess. Seeded accounts (alice/dana/bob, password `demo1234`) let you try the gate yourself. Note: This project is curated by the VibeFix editors from public sources. All rights belong to the original creator.

0

0

9

0

已上线

教育opencode

Moye

Timer-free learning for kids with ADHD or dyslexia: one spoken question at a time, a patient honey badger mascot, and difficulty set by plain readable code.

Moye is a free learning app for kids with ADHD or dyslexia — and it does the opposite of almost every learning app out there. Most learning apps run on a clock and a scoreboard: stall on a question and you get a red flash, a buzzer, a broken streak. Moye shows one question at a time, read aloud on request, with no timers anywhere. A wrong answer earns a hint and a patient honey badger named Moyin, and the score never drops. The build is for a friend: a 4-year-old also named Ayodeji, made for the Hacktoberfest Weekend Challenge "Build for a Friend". Under the hood the part the author is proudest of is that no model runs inside the app at all. Difficulty is decided by small readable functions — Bayesian Knowledge Tracing in `lib/mastery.ts` updates how likely a child is to know a skill after each answer, then moves them to an easier or harder tier. The 90 questions are validated as JSON with zod, 159 tests pass, and there's a `WHAT_IS_REAL.md` that explicitly states what isn't done (e.g. no efficacy claims). The accessibility details are real work, not decoration: a comfort panel for OpenDyslexic font, letter spacing, text size, calm motion and voice speed; voices in English, Nigerian Pidgin, Yoruba, Hausa and Igbo; a plain-language weekly summary for grown-ups; honey-drop rewards with a daily cap; and the whole thing works with Wi-Fi off, installs to a phone's home screen, keeps progress in local storage with zero third-party requests. Aligned to Nigeria (UBE), England (NC), US (Common Core), India (CBSE) and Australia (AC) curricula — and deliberately built with an AI coding agent on opencode, the same opencode challenge judges will recognize. Note: This project is curated by the VibeFix editors from public sources. All rights belong to the original creator.

0

0

9

0

已上线

音乐音频claude-code

Krisp

A calm browser beat studio with hardware feel: step sequencer, song arranger and full mixer — with Magenta.js in-browser AI to humanize and vary your beats.

Krisp is a calm, hardware-feel beat workspace that runs entirely in the browser — built for one producer friend during the Hacktoberfest "Build for a Friend" challenge. Make beats by hand on a 16-step sequencer (click, drag-paint, Shift+click for velocity, live record on keys A–K), arrange them into a song on beat and audio lanes, mix them through channel strips with a master chain (EQ, send, pan, fader, meter, reverb return, compressor, limiter), and export WAV and MIDI for your DAW. The timing is the thesis: everything musical runs on Tone.js in the browser with a 100 ms look-ahead scheduler and never waits on the network — the network is only for saving, the library and making new sounds. The AI is all open-source and in-browser: Magenta.js models humanize your beat (GrooVAE), generate variations (MusicVAE), continue it (DrumRNN) or morph it (MusicVAE), all undoable. Need a sound? ElevenLabs generates the exact one-shot or seamless loop you describe, and a hybrid search on Tiger Data (Postgres + pgvector) finds any sound by its name or by how it sounds — "warm dusty kick" actually works. The workspace holds 8 original preset beats (all credited as original to the project), autosaves with a save LED, and ships with 146 unit tests. And it's built the way we'd want every vibe project documented: every agent session was captured with Entire and linked to the commits it produced. The screenshots in the repo's "Blog images" folder — the song arrangement, the mixer, the ElevenLabs create flow — are the real studio, not mockups. A beat studio where "the groove is the 40 milliseconds a grid throws away" and it gives them back, one session at a time. Note: This project is curated by the VibeFix editors from public sources. All rights belong to the original creator. ---

0

0

10

0

已上线

效率claude-code

EchoNote

A local-first voice-to-action desktop app: press a hotkey, speak, and your words are transcribed and filed as a task, reminder, idea or note — all offline.

EchoNote is a voice-to-action desktop app built around a dead-simple loop: press `Alt + Shift + Space`, say whatever's on your mind, press it again and go back to work. It transcribes what you said and files it as a task, reminder, idea or note — "Remind me to call the dentist tomorrow" becomes an actual reminder with the date. And everything runs locally on the laptop: your voice, your notes and the AI processing never go to a server. The build story is peak vibe coding. The author doesn't know how Linux apps are made — his words: "claude made this app possible" — and the dev diary shows Claude Code doing serious engineering work. Speech-to-text is whisper.cpp, understanding is Google's Gemma 4 E2B running locally through llama.cpp, storage is SQLite, and the original transcript is saved before the AI processes it, so a note is never lost even if the AI is down. Then the agent kept going: a 2.7-second shortcut delay hunted down to 17–27 ms through a named pipe, Wayland's refusal to register global shortcuts solved with a GNOME custom-shortcut fallback, inference servers cleaned up when the app is killed, and 96 unit tests plus 8 real-model tests plus 7 end-to-end workflows driving the built app with WAV files as the microphone — including a full capture with the network interface down. First launch downloads about 1.1 GB of models from Hugging Face (CPU-only is fine, GPU acceleration is on the roadmap); after that, no internet needed. The demo screenshots show the real capture screen — a big mic button, the recent list filed into Reference/Idea/Task with transcripts, and "AI: loading…" honestly sitting in the sidebar. Made for the Hacktoberfest "Build for a Friend" weekend challenge, MIT licensed, installs in about five minutes. Note: This project is curated by the VibeFix editors from public sources. All rights belong to the original creator.

0

0

12

0

已上线

生活方式claude-code

Solar Dashboard

A home solar + electric-bill dashboard: per-panel live data, dying-inverter alerts, and a plain-English briefing written by a local AI every six hours.

It started as a son's attempt to prove he wasn't the reason Dad's electric bill was so high — and turned into a full home energy intelligence system. The Solar Dashboard pulls live data from the family's Enphase solar gateway every minute, down to each individual panel, combines it with ComEd electric bills and PVWatts government data on what the roof should produce, and ranks where the power goes per device and room via Home Assistant smart plugs. The standout feature is "Dad's Briefing": every six hours, a collector hands a small set of facts to Gemma 3 4B running locally on the home PC, which writes a short plain-English note at the top of the dashboard — no charts, just "your bill is $270.48 vs $46.12 last year, here's why, here's what to do." And the dashboard earned its keep: the per-panel map exposed that the solar company's layout didn't match the actual roof — panels reinstalled after storm repairs were never remapped, so the system was reporting panels in places they are not. The author proved it by checking where a tree's shade actually lands. Built with Claude Code (every session tracked with Entire, linking each AI session to the commits it produced), it runs Next.js 16 on Vercel with Upstash Redis, and even the bug stories are good — like the night the dashboard declared all 23 panels dead, because microinverters run on the panels' own power and go quiet after dark. Note: This project is curated by the VibeFix editors from public sources. All rights belong to the original creator.

0

0

11

0

已上线

AI 应用claude-code

Padh Ke Batao

Snap a photo of any document and hear it explained in simple Hindi or English — what it says, what to do, by when — read aloud, all processed locally.

"Padh ke batao" is Hindi for "read it and tell me" — what the author's grandfather says whenever he hands someone a letter he can't read. Padh Ke Batao is built for him: take a photo of a document — a bank notice, hospital report, pension circular, government letter — and get it explained in simple Hindi or English: what it says, what to do, and by when. Then it reads the explanation aloud, in a voice he actually enjoys listening to. The privacy architecture is the point. An open-weight vision model (Qwen 3.5) reads the photo locally through Ollama on a laptop with no GPU, so the photo never leaves the computer. A Fastify server extracts six structured fields — document type, summary, key details, action needed, deadline, urgency — with a deadline countdown, and only the summary text is sent to ElevenLabs for the voice; without a key it falls back to a voice installed on the computer. No account, no upload, no per-document cost. It was built with Claude Code as a pair programmer for the Hacktoberfest "Build for a Friend" challenge, and the dev diary shows the agent doing real work: renaming the package, regenerating screenshots when one was cut off, and reasoning through why ElevenLabs stays optional. One big button, Hindi/English switchable anytime, honest limitations section included. Note: This project is curated by the VibeFix editors from public sources. All rights belong to the original creator.

0

0

8

0

已上线

旅行本地claude-code

Drum Bun

A road-trip agent for Romania–Germany drives: cheapest vignette combos per travel day, dated rules, and every answer traced to the official source behind it.

Drum Bun — "have a good road", what Romanians say before a journey — is an agent that tells a driver exactly what their car needs between Romania and Germany or Austria. Ask "Brașov → Munich, 20 December, back 3 January, Euro 5 diesel" and it computes the cheapest set of vignette products per country for your dates, the rules that apply on those dates (winter tyres, emission zones, border checks), and the outdated claims you probably read online — each linked to the fact that corrects it. The clever part is the architecture: the content is modeled as structured pieces in Sanity (routes → legs → road sections, toll products with dated prices per Euro band, rules with yearly season windows, and claims with verdicts). A pure planner with 21 unit tests walks that structure — picking the price valid on each date, solving an exact set cover over Hungarian county vignettes — and the model only presents the result. The model never computes a price. On the author's 18-question eval in four languages, the agent scored 18/18 against 16 for semantic search and 4 for the same model with no documents. It was built with Claude Code for the DEV × Sanity challenge, and the build diary is unusually honest: the first eval run lost to keyword search (6/16), the knowledge base once inherited the author's own data error and hardened it into prose, and the author documents every fix. The live site works without login and includes a "How it knows" page showing every issue, both sides, and why one was kept. Note: This project is curated by the VibeFix editors from public sources. All rights belong to the original creator.

0

0

8

0

已上线

社交社区claude-code

Walkie

Push-to-talk voice app for Android and iOS: hold the button, talk, and your contacts hear you instantly. Real-time WebRTC audio with groups and presence.

Walkie is a real-time push-to-talk voice app for Android and iOS — think classic walkie-talkie, rebuilt for the smartphone era. Hold the button, talk, and your contact or group hears you instantly. Every participant joins muted and the microphone opens only while the button is held, with echo cancellation, noise suppression and speech-optimized audio keeping it crisp. Under the hood it's a serious piece of engineering: WebRTC audio over LiveKit for direct and group rooms, connection requests traveling over Supabase Realtime broadcast channels so the other person gets an "Incoming connection" sheet to accept or decline, online status via Realtime Presence, and a foreground service that keeps the audio session alive when the app is minimized. Contacts work by username or 6-digit Walkie ID, and security was clearly thought through — LiveKit room tokens are issued by a Supabase Edge Function that verifies the session, so no secret ships inside the binary, and all seven tables are protected by Row Level Security. The build story is the VibeFix part: it's a personal project by Antonio Pignataro, implemented end-to-end with AI coding agents (Claude Code) under his direction. He defined the product and the layered architecture and validated every feature on a real device — including a late iteration that moved LiveKit token signing out of the app and into the Edge Function. Light and dark themes included, and the README documents everything from the ES256 token flow to running it locally. Note: This project is curated by the VibeFix editors from public sources. All rights belong to the original creator.

0

0

12

0

已上线

商业

Puzzle Press

An AI agent built this puzzle-book generator and already made $19: word search, sudoku, mazes and crosswords as print-ready KDP PDFs, right in your browser.

Puzzle Press is the product of the bananafest-destiny experiment: autonomous AI agents building and selling software in public, with every number on the record. The agent behind it is called vibe-cider, and the site owner describes the whole thing, without flinching, as "vibe coded slop. We're owning it." Then a stranger arrived from a Google search at 2:44 AM, took a free book, made a free cover, and paid $19 for a price they were never shown until a dialog opened by itself. Net of Stripe's fee: $18.09. The scoreboard moved. The product itself is genuinely useful: a browser-based generator for Amazon KDP puzzle books. Pick word search, sudoku, mazes, criss-cross fill-ins or crosswords plus a trim size, and out comes the whole paperback — puzzles, solutions, page numbers and a matching cover — as PDFs that meet KDP's manuscript rules. Ninety word-search themes or your own word list, large-print option, fonts embedded, and your words never leave your browser. Free gets you a full-length book with one small footer line; $19 once removes it. It is a strange, honest artifact: a real tool, built by an agent, bought by a human, documented down to the timestamps. Whatever you think of the method, the $18.09 is real. Note: This project is curated by the VibeFix editors from public sources. All rights belong to the original creator.

0

0

7

0