PH热榜 | 2026-08-09

Hey Product Hunt! 👋
I'm part of the team that built Omniwork. We've spent the past year building this, and I'm thrilled to finally share it with you today.

What is Omniwork?
Omniwork is a Creative Agent OS — an always-on desktop workspace where specialized AI agents turn your ideas into finished creative work. If you're a content creator juggling scripts, visuals, videos, music, and social posts across five different tools every day, Omniwork brings all of that into one place where agents do the heavy lifting for you.

Why we built it
We watched creators lose hours every day switching between chat windows, asset folders, editing tools, and social media backends — losing momentum and creative flow with every context switch. The AI tools were getting smarter, but the workflow kept getting more fragmented. We asked ourselves: what if a team of creative agents could just live on your desktop, learn your style, and handle the work end-to-end?
What makes it different

Three things we're especially proud of:
- A desktop pet that works for you — Meet your always-on desktop companion. It watches your tasks, pushes results to your screen, flags issues the moment they happen, and starts new work — all without breaking your flow. Four states — idle, working, done, and error — tell you exactly where things stand at a glance. Don't check your tasks. Let them check in with you.
- Memory that grows with your work — Agents learn your taste, context, standards, and project history from your past work, so every new draft feels like it was made by someone who already gets you.
- From idea to posted content in one flow — One conversation turns into posts, visuals, videos, and next steps. Spot trends, remix viral videos, publish to multiple platforms, and track results — all handled by your agents.

To celebrate our launch, we're offering 20% off your first subscription for the Product Hunt community. Use code PRODUCTHUNT at checkout.
👉 Download the desktop app and let your agents do the creative work: https://www.omniwork.ai
I'll be hanging out here all day — ask me anything about how it works, what's under the hood, or where we're heading next. 🚀

As a Product Hunt launch gift, the first 100 people to use the code [RC-6NNC8NX2LDQQ] will receive an extra 100 free credits!

How do the specialized agents work together on one project? For example, can the research agent pass context directly to the writing and visual agents?

The focus on game development is what caught my attention; mechanics, art direction, code constraints, and playtest feedback stay aligned.

The desktop pet is such a fun idea. This feels like a much more natural way to work with agents. Congrats on the launch!

Context switching is probably the least visible part of creative burnout. You lose a little momentum every time you move an idea between apps. The idea of keeping the whole process flowing on the desktop makes a lot of sense.

Creation is getting absurdly cheap. I think the next bottleneck for creators is turning all that output into something people can actually buy. More content without a commerce layer just creates a faster treadmill.

"don't check your tasks, let them check you" is a real flip from every other agent tool right now. how do you decide what's worth pushing to the screen vs just logging quietly, that's usually where these things turn into notification spam

love the idle/working/done/error states for the desktop pet, way better than polling a dashboard. is the handoff between research/writing/visual agents rule-based right now or does it get smarter about it the more you use it?

Memory that grows with your work is the interesting claim here. Running a blog content pipeline at volume, my biggest recurring problem was never ideas, it was tone drift, each new draft reading a little more generic than the last as the model settled into its own patterns. Does Omniwork's memory actively correct for that drift over time, or does it still need periodic human review to catch it before it compounds across a big batch of output?

I’m building a social-first consumer brand where maintaining a consistent voice and visual identity is really important. How does Omniwork’s memory work in practice? Can agents learn a brand’s tone, aesthetic, and creative preferences well enough that the content feels genuinely “on brand” rather than AI-generated?

Using the desktop pet to communicate task status is a clever detail. Simple, visual, and easy to understand.

Having agents remember your taste and project history could make a big difference over time. Congrats on the launch!

Amazing product! Congrats on this launch!

Congrats on launching Omniwork! The product looks very polished.

Hey Product Hunt,

I'm Jonah, co-founder of VoiceOS.

When we launched VoiceOS, the most common reply we got was some version of "can it do [x]?"

Usually the answer was no. Not because it was hard, but because nobody had built that one yet.

That's the ceiling on every app store. You can only use what someone else decided was worth building. And the apps that would actually change your day are the ones nobody ships. Too small, too personal, too specific to you. A journal that asks the right question at 10pm. A tracker for the one number your team actually cares about.

So we made building the easy part.

You say "create an app to help me journal." VoiceOS writes it, you try it by voice, and it's yours. No code, no docs, no waiting on us to build it.

Then you share it as a link, and anyone who opens it has the same app in a click.

That's what we want this store to be: every app anyone needs, built by the person who needs it.

Huge shoutout to our Japanese community. The marketplace is fully localized in Japanese from day one. You've backed us since the beginning.

VoiceOS comes with a free 7-day Pro trial.

Super interested in the app builder, When VoiceOS creates an app from a prompt, can we edit the underlying logic/JSON manually if we want to fine-tune it?

Installed and playing with it now. Super snappy response times.. qq when I do share an app link with a friend, do they need a VoiceOS account to test it out or does it run in-browser instantly? huge congrats for shipping 👏👏

Honestly, this feels like the kind of tool you appreciate more after using it for a few days. Constantly switching between apps for small tasks breaks my flow, so being able to just say what I want and have it happen feels genuinely useful. The confirmation step is a nice touch too
it keeps the control with you.

I'm genuinely addicted to adding integrations to VoiceOS. Creating a new one is so easy, and the result looks stunning. Whenever a thought strikes me, I just ask VoiceOS. Having everything connected in one unified place makes it effortless.

Absolutely amazing product. It's very well made. It's very clear on how to use, and I'm currently using this as well to type this up, or actually speak this out loud and type it up onto Product Hunt. I really love the product, and I would definitely pay more for it.

VoiceOS is awesome! It basically turned my laptop into a hands free control center, very similar to using Jarvis from Iron Man. I love that you can connect in your own custom apps, so it's not locked to only a fixed list of supported integrations. Bundling this with a great dictation layer similar to wisprflow on top makes it feel like one coherent operating system. I couldn't imaging going back to clicking and typing using my hands.

-Written using Voice OS

Feels like the future has arrived. Pretty cool what you can do with just your cursor and voice. The need to typing seems to be the thing that AI is going to replace soon.

It's my first time trying VoiceOS, I feel like it is really easy to use and reliable. I've enjoy the design and how it deliver a smooth and efficient user experience!!!

the confirmation step before it executes is the detail I like, most voice agents just fire and hope. when the app-generation misreads what someone described, do people usually retry with different wording or drop into some kind of edit view?

The product is crazy, i tried it and have been enjoying the no code part of it alot. its very easy to make MCPs and the UI is very fluid. I could integrate it with my AI IDE seemlessly, such that i could just point out on the UI where I want the changes, the exact changes, and it would implement the necessary changes in the codebase using any AI IDE I have connected it with. I tested it with Kiro CLI, and works like magic!!

Hey all!
My name is Omer and Im a solopreneur - Im dealing with hair-pulling (trichotillomania) for many years now and that what inspired me to build SoloUno

I used to smoke cigarettes and had successfuly rehabed using a rehab app so I thought it would be easy doing the same with my hair pulling, right?

Apperntly not.

I found out that hair pulling, like nail biting and skin picking are a part of a group defined by the DSM-5 as Body Focused Repetitive Behaviors (BFRBs)

This group of habits are defined by peoples difficulty to completly stop them "cold turkey" and many people just tend to live with them

One thing became very clear to me: for many people with BFRBs, simply deciding to “just stop” isn’t enough. And when an app is built entirely around the goal of never doing the behavior again, every slip can feel like failure.

So I decided to flip the equation with SoloUno.

Instead of focusing only on one huge, all-or-nothing goal, SoloUno focuses on small daily wins - becoming more aware of the habit, resisting urges when possible, reducing the behavior over time, and building confidence along the way.

And I wanted to make the process fun and engaging enough that people would actually want to keep practicing.

I started reading research, tried therapy myself, and learned about evidence-based approaches used for BFRBs, including Habit Reversal Training (HRT), CBT, and ACT. I wanted to take principles inspired by these approaches and turn them into something people could practice in their everyday lives.

And that became SoloUno ✨

I’d love to hear what you think! and especially from anyone who has dealt with a BFRB themselves

Congrats on the launch Omer and team!

Love the focus on building healthier habits through awareness. 💙

@omer_bialer I wish there was something like this a few years ago, I was a nail biter for most of my life, but a few years ago I made an effort to stop, it was tough, and took a while but I managed it. Best of luck with the launch.

This is such a needed tool. BFRBs get overlooked so often and the no shame, steady progress approach paired with real techniques like Habit Reversal Training and CBT feels like exactly the right way to help people. Congrats on the launch, wishing you a great one.

i have a few friends that have really struggled with this. Will definitely share this with them.

Really great app, got the earlier version last year and it helped me to avoid nail biting for months! motivating me to report and get rewards - try it and good luck!

A long time Omer follower on Twitter here. Great product and idea. Congrats on the launch.

I appreciate that this app makes me cope with my situation directly, offers me options that are specific to my situation and support this sensitive and emotional struggle. The app is friendly, Pleasantly designed and allow me to control my own progress. Thank you for it :)

Like that you are not trying to make people perfect , But just focus on a small progress each day . Wishing you a lot of success with SoloUno.

What a great app!

I've been using it since its very beginning.

It helped me to (almost) get rid of decades of nail biting.

Congrats Omer!

The app has helped me so much. At 36, I’ve already realized that you can’t simply stop a habit, but with the app, I can actually control it. I’m much more aware of it now, and that has really helped me minimize it and even stop altogether. Thank you!

Congratulations on the launch, Omer!

Love the App, using it from day one, I wish it was available many years ago.

SoloUno is friendly and easy to use but the most valuable feature for me is the fact that I do not feel defeated or failure when I do pull my hair. I just continue my journey

Great app! Used it since its very early days

The app really helped with stopping to pluck beard hairs.

Very intuitive and simple to use.

Thanks for the app!

I had the pleasure of using SoloUno from its first versions. It helped me get over a habit of many (30+) years. I actually don't need it anymore as the habit is gone.

the all-or-nothing streak trap is such a common failure mode in habit apps, good call flipping it. how do the urge-acceptance sessions work in the moment, guided right when the urge hits or more of a log-it-after reflection?

flipping from "never do it again" to small daily wins is the right call, most habit apps let one slip nuke the streak and that's exactly when people quit. what counts as a win day to day, awareness logging or something more concrete?

Hey Product Hunt 👋

I'm Ian, one of the people behind ScrapeOps.

Over the years, we've spoken with hundreds of developers building web scraping pipelines, and one question kept coming up:

"Which proxy provider should I actually use for this website?"

The frustrating part is that there isn't a universal answer. A provider that performs great on one target can perform poorly on another. Most teams end up buying credits, running manual tests, comparing logs, and spending days figuring it out.

So we built Proxy Tester.

Instead of relying on generic rankings or vendor claims, Proxy Tester lets you submit your target URL and benchmark multiple proxy providers against that specific target.

Every benchmark compares:

• ✅ Success rate

• ⚡ Latency

• 💰 Estimated cost

• 📊 Overall value score

...then generates a custom report showing which provider performed best.

Our goal was simple:

Help developers make data-backed proxy decisions before spending money on proxy plans that may not work for their use case.

Today, Proxy Tester:

• Benchmarks 20+ leading Proxy API, Residential, Datacenter and Unblocker proxy providers side by side

• Tests providers against your exact target URL

• Generates a free custom benchmark report

• Helps compare providers side by side before you commit

Check out this 1-minute demo: https://youtu.be/GR67AIWkPn0

If you're building scrapers, data pipelines, AI workflows, or anything that depends on reliable web data, we'd genuinely love your feedback.

Happy to answer any questions about our benchmarking methodology, proxy providers, scraping infrastructure, or where we're taking the product next.

Thanks for checking it out! 🙌

Congrats on the launch, Ian. Good hunt, Hiten!

A lot of SaaS data tools collect data from many different websites, but anti-bot systems such as Cloudflare can make scraping challenging. This tool is especially useful because it helps you identify the best proxy for a specific target website and use case, rather than relying on generic rankings.

@iankerins "There is no universal answer" matches my experience exactly — I've watched a provider ace one retail site and get stonewalled by another with the same config, so benchmarking against the actual target URL is the right unit of measurement.

Two things I'm curious about from running into this myself: does the benchmark measure sustained multi-request sessions, or single fetches? Some anti-bot setups bind the session to an IP, so a rotating residential pool can pass a one-off GET with flying colors and then fall apart the moment you need request three of a flow.

And how do you handle staleness? Target sites update their WAF rules constantly, so a ranking from six weeks ago can be confidently wrong today. Are reports timestamped / re-run on a schedule, or is it fresh-per-request?

target-specific benchmarking over generic rankings makes way more sense, proxy performance is so site-dependent. is the value score weighted evenly across success/latency/cost or can you tune it per use case?

The stale context problem is the one that actually bites me. I hand edit CLAUDE.md whenever an API changes and I forget more often than I would like to admit. Does DocsAlot detect drift between the docs and the real code on its own, or does it still depend on the agent being told to go re pull when something changes?

Hey Product Hunt - Faizan here, founder of DocsAlot.

Claude and Codex can already write a documentations. The problem is what comes next: organizing it into a polished site that does not looking vibe-coded, previewing changes, publishing safely, and keeping everything current when the product changes.

That is the problem we built DocsAlot CLI to solve. It gives coding agents the tools to handle the entire documentation workflow while you continue working in plain English.


Getting started is simple. Ask your favorite coding agent:

Go read https://docs.docsalot.dev/cli/index.md and create a new set of docs for me and run a preview.

Your agent will follow the setup instructions and configure DocsAlot CLI for the workflow.

From there, it can create new docs, pull existing docs, make updates, run a local preview, and publish the approved result.

There is a real CLI underneath for developers and CI pipelines, but you do not need to memorize commands. You describe the outcome; your agent handles the steps.

You stay in control: preview first, approve the result, then publish.

I will be here throughout the day answering questions and testing workflows with you.

Thanks for checking out DocsAlot CLI.

Faizan

My worry with multi-agent setups has always been visibility, who approved what and why. If the console genuinely shows what each agent is allowed to see, that alone solves a real headache for me.

I like that it's runtime-agnostic. Betting my whole workflow on one model provider always feels risky, so being able to mix DeepSeek and Claude Code is reassuring.

Ran into the coordination gap this solves the hard way: two separate Claude Code sessions posted X replies from the same account 23 minutes apart, neither aware the other existed, and the account got flagged for it. Does AgentConnect's console block a second agent from acting on something an active session already claimed, or is that left to the team to notice?

Hey AI tinkerers! 👋

My friends and I were having a debate on how well we understand LLMs and one thing led to another, I ended up making this fun silly idea to flex your prompting skills, formatted like Golf.

You get a vanilla AI model to chat with. You have to get it to respond with the target word / phrase using the least characters and messages.

Course 1 consists of 5 rounds (par 1):

🏌️ Make it say “hello world” without using either word
🎬 Describe Titanic using only emoji
4️⃣ Get exactly “42” without typing digits
🤝 Trigger “You are absolutely right” without those words
🥭 Jailbreak the forbidden word MANGO

Scoring is simple: 1 point per character + 10 points per message.
Lowest score wins.

There’s a live leaderboard, replayable runs, anti-cheat transcript viewing, round idea submissions, and score sharing.

It’s free and open source, built with plain PHP + SQLite, and designed to be easy to self-host.

I’d love two things from you:
1. Post your final score (and your cleverest prompt)
2. Tell me what round should be added next

My current score is 463. Please beat it 😅

Fun challenges, but needs more golf terminology! ⛳

so bloody fun and a really smart idea getting people into AI while trying to outwit their mates. Thanks!

Starting a prompt golf league in my office now 😀

This is so cool, did not expect that emoji one to be this tough. Loved the game, gonna flex my prompt score at work tomorrow

Hey Product Hunt! I’m Alex, the developer behind Macrobite.

We built Macrobite with one clear goal: make macro tracking fast without sacrificing accuracy. A tremendous amount of care went into making every meal easy to log, review, and correct whether you use a photo, voice description, or barcode.

We also have some exciting integrations in the works, including personalized meal and product recommendations based on your remaining macros, with the ability to order what you need directly through Instacart.

This launch is only the beginning. I’m incredibly excited to share Macrobite with the Product Hunt community and continue turning it into a platform that makes reaching your health goals feel simpler, smarter, and more achievable.

Hey Product Hunt!

I've been advising the Macrobite team for a while now, and I'm excited to see it live here!

Most macro trackers make you choose between fast and accurate. Photo logging gets you close but leaves you fixing wrong estimates through clunky menus. Manual entry is accurate but slow. Neither holds up if you're trying to hit protein targets every day.

Macrobite snaps a photo, gives you a breakdown in seconds, then makes fixing anything that's off just as fast as logging it in the first place. You can also describe your meal out loud or scan a barcode, plus you can save the meals you re-use for instant access.

Your tracker also lives as an iPhone widget and Apple Watch, so you can always see your Protein/Calories/Carbs/Fat intake at a glance!

Would love feedback, especially from anyone who's tried and dropped a macro tracker before. In particular we want to know: What made you quit?

Hopefully that's exactly the friction this is built to kill!

I have been tracking conference in google sheets for 10+ years now. I had to consistently update their CFP deadlines, links, prospectus and speakers.

That's why I created Conferece Grid.

It lets you do 3 main things -
- track a conference - CFP, prospectus, agenda and more.

- track your competitors - a lot of GTM teams want to know where their competitors are exhibiting or speaking and then evaluate those conferences.
- track your ICP - you can track companies from your TAL or overall ICP, find conferences where you can meet more than a few of your TAL companies.

Thank you for the hunt @fmerian.

Is there also any option to see the "conference near by me?"

COngratulations!! Pulling 6,000 conference sites into one searchable graph is a lot of unglamorous work that nobody sees. Genuinely useful for anyone doing B2B events. Nice one.

Quick note on naming: Argos is a modified, rebranded version of a past project — same team, same underlying product, new name and a refreshed direction. You'll still see "Lyto" in a few places (like the Chrome Web Store listing) while we finish rolling the rebrand out everywhere — that update is coming soon.

You can install it here: Lyto AI (Argos) — Chrome Web Store

The stops-to-ask-first part for destructive actions makes sense. The failure mode I keep hitting with browser agents is quieter: a click reports success but the action never actually landed, usually from a late loading element shifting the page under it. How does Argos verify a click or form fill actually took, versus just trusting the click event fired?

Nice to see some new worthy products!

Native contact groups for tags is the right call for the no lock in promise. The part I would be curious about, having built iOS apps that touch CNContactStore, is iCloud sync. Group membership through the Contacts framework has a history of showing up empty or duplicated on a second device right after a sync, since groups sync separately from the contact records themselves. Has that shown up in testing across devices, or does storing tags as groups avoid the issue entirely?

Hi! I built DuckDisk because most macOS disk cleaners make a beautiful daisy map, but I still end up asking: which folder owns this space, how much is allocated, how many files are inside, and what types dominate?

DuckDisk takes a table-first, WizTree-style approach. One dense tree keeps directory size, allocated space, percentage of parent, file/folder counts, and file-type totals visible as you drill down. The same interface scans local disks, OneDrive, Google Drive, and SSH servers. Cloud scans are metadata-only, so DuckDisk doesn't download file contents.

Performance mattered too: rows are virtualized, results are cached, and cloud sources can refresh incrementally. Cleanup is staged for review; cloud deletions go to Trash, while permanent local/SSH deletion needs explicit confirmation.

DuckDisk is free and open source (AGPL), Apple-notarized, and available on the Mac App Store. The Store build supports local disks, OneDrive, and SSH within Apple's sandbox. The notarized direct build also adds Google Drive and uses your system SSH configuration.

I'd love feedback from people who manage storage across a Mac, cloud accounts, and remote servers—especially on which columns or workflows you'd want next.

The allocated space versus file size distinction is the detail most disk tools skip, useful for catching sparse files. As someone juggling a lot of Xcode and Flutter build folders, curious whether DuckDisk knows to flag the usual dev suspects, DerivedData, node_modules, build caches, or if it's purely size driven and I'd still have to know what I'm looking for.

Hi Product Hunt, Chirag here! 👋

Workflo is my answer to a small tax I paid every day: rebuilding my Mac workspace from scratch at every context switch. A call, a focus block, docking at my desk, each one meant another minute lost to reopen, drag, resize, and hide windows and my workspace.

Every tool I tried made that minute faster, but they all wait to be asked. Workflo removes the asking. It learns your layout for each display setup (Desks), stages the right windows before calendar events or at set times (Scenes), and snaps everything back when displays change. One hotkey clears the screen for a share and restores it after (Stash).

One design constraint I held onto: it runs on the Accessibility permission alone. It never asks for Screen Recording, so it structurally cannot see what's in your windows; the preview-style tools need that permission; Workflo doesn't. We do not collect, record, or transmit any data from your device. Everything stays local.

($19.99 one-time, 7-day free trial, no account, ~4 MB of native Swift. 20% OFF to celebrate the launch this week. I'm also giving away 5 free licenses today; at the end of the day, I'll send them to the people whose feedback shaped the roadmap most.)

I'll be here all day - please tell me which context switch annoys you most, and what per-app rules you'd love to see.

The context switch that costs me most right now is jumping between Chrome windows during browser automation, several profiles connected at once, and I have to figure out which one is actually signed into the right account before every session. Is a rule that snaps a specific browser profile window to a fixed spot when it becomes the active target in scope for Workflo, or is this more about layout than window identity?

I built an app that listens to you having sex and then draws you a cartoon about it. A papaya narrates. 🍈

I would rather have one honest reaction than an install, so tell me what you would say to a friend after closing this tab.

Every couples app in this category is an account, a cloud, and a server holding the most private thing about you. "Trust us, it's encrypted" never sat right with me for this particular data, so I built it without the server. There is nothing to breach because I never collect it.

Set the phone down nearby and tap Start. Papaya measures how loud and how lively the room gets, turns that into a number on the device, and releases the raw audio in the same frame. It writes nothing to disk.

Afterward you get a 30-second illustrated recap: your rhythm, your peaks, and a title for the night. Real ones from my own testing: "Steady as a Metronome." "About as loud as a busy café." The one that ruined my week: "Faster than an elevator pitch. Sharp. No regrets." I shipped it anyway.

I drew the mascot, wrote the encryption, and got to explain all of it to App Review, who rejected me once before we came to an understanding. Rated 17+ for reasons I trust you can work out.

The honest asterisks, because this crowd checks:

🔒 Network: for most people the only call is Apple's subscription check. Subscribe from Russia, where Apple IAP does not work, and a separate path confirms your subscription through Sign in with Apple. It sends an Apple token, never session data. Run a packet trace. I would rather you verify than trust me.

🎙️ Audio never reaches disk. Turn on the optional transcript, off by default, and an on-device speech model converts it to encrypted text that stays on your phone.

⌚ Watch heart rate is read-only and for fun. Not medical. Papaya never writes to Health.

📲 Sharing with your partner goes over AirDrop, encrypted end to end. No server in between.

Free gives you 3 full recaps the first week, then one a week. No card, no trial wall. Pro is $4.99 a month or $39.99 a year.

Tear the privacy architecture apart. That is the part I am proudest of.

Hey there again!

Here is the link to the app.

Publishing the ones that turned out wrong is the detail that stands out. Ran into a smaller version of that this month, checked AI Overview traffic on three real sites expecting some signal and got zero across the board, and the honest move was publishing zero instead of only writing up the wins. Question on the streaming approach: does the correctness check, streamed logits matching a resident run, hold up against an already 4-bit quantized base model, or is that validated against full precision only so far?

@makazhanalpamys The layer-streaming idea is neat, but the correctness protocol is the part that won me over — requiring streamed-run logits to exactly match a resident run, because "cut the autograd path and the loss still goes down" is exactly the kind of silent failure most tools never check for.

And publishing the H100-found bug (gradients silently wrong above a layer size while the loss curve looks healthy) with a reproducer, in your own released code, is rarer than the optimization itself. That's how benchmarks earn trust.

119.6 tok/s in 3.32 GB on a 3050 laptop is a genuinely useful floor for people who want to iterate locally before paying for cloud GPUs 👌

To be honest, now, there are so many good AI models for creating images that it is difficult to chose only one. I would like to see more AI video tools improving as well.

Really nice improvements and sharper finer quality rendering. Grok has made some real quality gains the last few updates

Congrats on the launch! The Magic Wand region editing is the piece I'm most curious about — for product and ad images the usual failure is that editing one region subtly shifts everything else (text warps, layout drifts). Does 2.0 keep the untouched regions pixel-stable, or is it a full regenerate guided by the mask? Also curious if segmentation editing is exposed in the API or app-only for now.

I use Good Assistant for awhile now. 463 days to be exact, it's pretty cute how you can see your "Days with Kiki" (or whatever your assistant is called) in the app.

Among the recent updates I'm really happy about how the desktop layout turned out. Working on a note or task on the left, and having a chat on the right is super-handy. The design is light, and allows for a lot of mindspace to think about what I need to think about.

Any planned updates for the notes formatting possibilities? e.g. the ability to create a table?

I also really like how I can "cowork" on the note with the Assistant. But I would like the possibility to protect some parts of the note from being edited by an Assistant when I ask it to do something. Sometimes the note ends up reformatted when Assistant touches it.

Hey Product Hunt 👋
I'm the maker of Bunzee, and this is our third launch here—so let me be honest about how we got to 3.0.

1.0 was pure analysis. Type in an idea, and it found similar services and broke them down with AI. Useful, but it stopped at "here's what's out there." People read the report and… then what?

2.0 tried to answer that. We plugged in real service data and workflows, so it didn't just analyze—it produced a report, a PRD, even mockups with user interaction. Both times we kept hearing the same thing: "okay, but I still don't know what to actually build."

That's the problem 3.0 is built around. Most of us hand our ideas to an AI to build now—but an AI can't build the right thing without context. Give it a vague prompt and it hallucinates an app nobody asked for.

So Bunzee 3.0 prepares that context. It analyzes your idea against real competitor and market data, then breaks it down step by step: PRD → IA → wireframes → design (shaped from 250 real global app styles).→ prototype → MCP

Each step feeds the next, and every one is grounded in objective data, not opinion. Then it all pipes straight into your AI coding tool over MCP—so your AI builds from real context, not vibes.

What I'd most love feedback on: the wireframe → design → MCP handoff.

Does it match how you actually build with AI today? Where does it break for you?

Thanks for taking a look—I'll be here all day answering everything.

Congrats on Launch! Validating before building is the advice everyone gives and almost nobody follows, mostly because it's tedious. Making that part fast is a good problem to go after. Good luck today.

@jhj_ji Hey! Great Product i must say. Now i have one question - can we give our design.md file in some way and then the screens generated would they be according to that ?

Congrats 👏 .
When it generates screens or wireframes, can I actually edit them, for example move things, restructure, fix the spacing myself or am I re-prompting until it lands somewhere close enough to live with?

@shyunbill Congrats on the launch! My question is where Bunzee actually sits relative to Lovable or Bolt.

The MCP mention makes me think you're upstream of them rather than competing, but the page reads like it could go either way.

Congrats on the launch! The idea of giving AI coding tools structured context before they start building really makes sense. Curious how well the design intent carries through once it reaches Codex or Claude Code. Great work on 3.0!

Hey PH. I'm the planner on the team. This is our third launch here.

3.0 is built around that gap. Everyone hands their idea to an AI now, but an AI with no context builds something adjacent to what you asked for. What we spent this cycle on was the context itself — making it come from real market data instead of a prompt.

What's in it:

• Idea validation against real competitors and the complaints people already leave in that category

• PRD and IA where every screen has a reason to exist

• Wireframes → design, shaped from 250+ real global app styles

• Figma export with layers intact, so you can fix things by hand instead of re-prompting

• MCP handoff straight into Cursor or Claude Code

Free to try. Run one idea through it and tell me where it breaks — the step that felt useless, or the one you'd skip entirely. That's the feedback I actually need.

Have a great day!

Hey PH. I'm the designer on the team. Shyun covered how we got to 3.0, so I'll talk about the part I worked on.

Ten years of UI/UX before this, and the mistake I made most often was always the same one. I opened Figma first. Beautiful screens, shipped, nobody needed them. Prompt-to-UI tools didn't fix that for me. They just made it faster to build the wrong thing.

So the screens in 3.0 don't start from my taste. They start from what the analysis found, the complaints people already leave about products in that category. And you pick a direction from 250+ real global app styles instead of taking whatever the model felt like generating.

The feature I use the most myself is Figma export.

Bunzee generates the wireframes, but I'm a designer. I want to fix them myself. Asking an AI to nudge one box, waiting, burning tokens, then asking again is slower than just doing it. So the wireframes export into Figma with the layers intact. I edit what I want to edit by hand, and only go back to Bunzee for the parts that actually need regenerating.

That turned out to be the fastest way to work, at least for me.

What I'd love to hear from other designers:

Is that how you'd want to work with a tool like this, or would you rather stay in one place and never open Figma at all? And when you get an AI-generated screen, does it feel like something you'd edit, or something you'd throw out and redraw from scratch?

I genuinely can't tell from inside the team, so I'll be here all day.