PH热榜 | 2026-07-03

Hey Product Hunt 👋

I'm Thomas, co-founder and CEO of Raycast. Today we're opening Glaze to everyone, and I think it's a small step toward software finally getting personal. We couldn't be more excited to share it with the Product Hunt community.

Here's the thing that got us here. Tools are how we make progress, and today so many of our tools are apps. But most software is a compromise. To reach millions of people, you build for the average, and the average doesn't really exist. We're all individuals, so the "one app for everyone" always leaves something on the table.

Glaze flips that. You describe what you want, and it builds a Mac app shaped around you. It lives in your dock, launches instantly, works offline, and taps into the full power of your machine. When something isn't right, just talk to it and change it. It's so much fun iterating on it to get exactly what you're looking for.

And it doesn't stop at building. You can publish your app to the public store for anyone to use, or share it privately with your team, so internal tools stay internal and good ideas spread. We run our support and sales processes fully on Glaze apps at Raycast. Additionally, we've seen teams like Cursor, Linear, or Vercel building custom apps to optimize how they get work done.

A few of my favorite apps:

• World Cup: Simply stay on top of all matches, with live stats, a beautiful knockout stage view and more.

• Yolo: A terminal that is optimized for using agents with vertical tabs, AI suggestions and a ton of keyboard shortcuts.

• CS Glaze Synth: A full synthesizer 🤯 The app was used to create the sounds for our launch video.

Glaze is free to start, and there's a paid plan when you want to go further. Mac first, with more platforms to come.

Honestly, the part that blows me away is the creativity. People keep building things we'd never have thought of, and that's exactly what we hoped for.

One thing I'd genuinely love your take on: what's the one app you've always wished existed but never had the time or setup to build? Sign up, try to build it and share your link below.

Thanks for taking a look. We're early, and we'll learn most of this in the open with you 💠

Hey everyone 👋

I’m Alex, Product Designer at Raycast and part of the team behind Glaze. We’re incredibly excited to finally share Glaze with the world. It’s been a wild journey turning an idea into a product that lets anyone build fully functional, beautiful desktop apps.

A few of my favorite things about it:

• Describing an idea and watching it appear in my macOS dock, ready to use anytime

• Using Annotation to make changes directly inside my app and watching them happen in real time

• Generating app icons and exploring different concepts until one just feels right

• Browsing the Store and discovering apps built by my team and the wider Glaze community

A few apps I built myself:

1. Hotkey Explorer: helps me explore and organize hotkey systems for Raycast

2. Winamp: brings my favorite music player from the 2000s (and all its wild skins) to macOS

3. Glassmaker: helped me design the Glaze app icon by exploring glass shapes and materials

We can’t wait to see what you build! Share your apps and let us know what you think. We’re just getting started, and your feedback will help us shape what comes next for Glaze. 💠

Hey, one of the makers of Glaze here 👋 Being able to build truly personal apps in just a few prompts is SO powerful, especially when you don't have to think about boilerplate setup and how to distribute the app with others.

Here's a few personal ones built for myself:

• Spotify Classic: A dead simply Spotify client inspired by the original macOS app, no fluff - just playlists and search.

• Avalanche Map: Visually plan ski touring adventures with 3d maps showing the latest official Swedish avalanche report.

• Shader: I'm a huge sucker for evening sun, this makes it easy to visualize shadows from mountains, buildings and trees at any location, date, and time.

• Ray.fm: Create your own radio stations, and vibe to the retro cassette-style UI.

• Funky Mirror: Play with fun webcam effects on your laptop.
  

Really excited to see what everyone is shipping on the Glaze store!

Amazing tool!

I'm not a developer, but I made an app for myself in very short time — a diary writing app. Now I finally have what I've always wanted. It really blew me away how easy it was to make.. Glaze isn't just for developers; it's also for regular users like me. I already have several more apps in mind.

Congrats on the launch! 🎉

I jumped in early and already built two apps to test it out. Going from “I wish this existed” to something sitting in my dock in a few minutes still feels a bit magical.

To answer Thomas’ question: the annotation flow is what sold me — being able to just talk to the app and watch it change in real time made iterating genuinely fun rather than a chore.

Excited to see where this goes. 💠

Hello fam!

I built a Glaze app to help me archive all of my analog scans. It's such a personal app that I never would've found one that worked the way I wanted it. I feel happy every time I use it 😁 I even made a video on my YouTube channel about, check it out to see it in action.

I'm also currently building an app to help me manage everything that's in mymind!

The interesting constraint here is that "create a Mac app by chatting" covers a huge range of outcomes, from a quick menu bar utility that does one thing to something with persistent state and real UI. Curious where Glaze actually sits on that spectrum. Specifically, what happens when the generated app needs to store data between sessions or talk to an external API, does it scaffold something real that you can open in Xcode and extend, or is the output more like a self-contained script that lives and dies inside Raycast?

Glaze is super fun — kind of the dream of site-specific browsers brought to the generative AI era!

I've built several personal apps so far and published Tesla Energy to help me track my solar production!

Hey everyone 👋 I'm Pavlo, AI engineer on Glaze.

I build the AI agent behind Glaze, from writing its prompts and refining the harness around it to wiring everything together so it can plan, write code, and validate its own work.

My favorite part is seeing all the effort and engineering behind creating an app disappear into an experience that feels like magic. You describe the app you have in mind, and Glaze helps bring it to life.

On a personal note, beyond being a blast and an awesome challenge to work on, Glaze has also had a real impact on my day-to-day life. I used it to build something I genuinely needed, a dashboard for monitoring my glucose levels tailored to my needs. I never felt like the existing options gave me exactly what I wanted, so I built one for myself with Glaze.

Can’t wait to see what you build. Share your apps, send us your feedback, and help shape what comes next for Glaze 💠

I've seen early previews of Glaze and it has been really impressive. Just like everything else from the Raycast team, the level of polish and attention to detail has been amazing. Looking forward to building my own mini-apps without having to do any of the heavy lifting that is usually required for this!

Love Glaze! Built so many apps for my personal workflows and some fun ones too. Latest is a tron inspired snake game.

I've tried Glaze earlier this week. Would've been great if we could export apps as xcode projects. Other than that, it works really well. Love it!

I like the idea of Glaze, but are there any plans to let us use our own AI subscriptions instead of Glaze’s credit system? That’s my biggest concern with AI tools at the moment, as nobody likes paying API pricing, but I would happily pay for “Pro Features” on Glaze.

👀👀👀

the free allowance is 120 credits, how much is it actually?

what is the biggest desktop app you can build from scratch on the free trial?

same question for the paid plan, because the cheapest paid plan is $20 which is same as Codex or Claude Code entry subscription price, and this is what Glaze 200 credits compete with

congrats with the launch, it's gonna solve so many quick problems!

Tried it yesterday and already love it. Made a little synth in one prompt, then refined it with three more — now I’ve got this cute vintage synth.

Works really well, and the publish flow is super smooth.

Congrats to the awesome Raycast team 🎉

Hey PH 👋 I’m Alexandr, one of the engineers on Glaze.

One of the most fun parts of building Glaze has been dogfooding it. I spend a lot of my day juggling git worktrees, dev servers, and multiple AI coding agents, but no app felt quite right for my workflow. So I built my own terminal app with Glaze: Yolo.

This is why I’m so excited about Glaze: it makes software feel personal. Not “one app for everyone,” but apps shaped around the exact workflows, habits, and preferences of the people using them.

Can’t wait to see what you build with Glaze — please share your apps with us 🙏

The idea of building Mac apps through chat sounds really approachable, especially for the vibe coding / AI workflow automation crowd. How much control does Glaze give over the final app behavior after the initial chat — can users keep refining the app in conversation, or is it more of a one-time generation flow?

The idea of creating Mac apps through chat sounds useful, especially for people who have a workflow in mind but don’t want to start from a blank coding setup. How much control does Glaze give over the final app behavior once the AI generates something? For example, can users keep refining the app through chat, or is there also a more developer-style way to inspect and adjust what was created?

"to reach millions you build for the average, and the average doesn't really exist" is going to live in my head, I run a niche workspace product and this is the exact argument for why vertical software beats horizontal, just taken one step further down to the individual

how does it handle updates when you tweak the app later, do you just re-describe the change and it patches the existing build or does it start from scratch each time?

Been messing around with Glaze for a few days. The idea of describing an app and having it show up in my dock is pretty wild. Built a couple of tiny utilities for myself without writing a line of code. It handles connecting to Notion and GitHub surprisingly well. Still early, and complex apps might be a stretch, but for quick personal tools, it's genuinely fun and useful. Feels like the future of making software personal.

Cool product! I tried to make a music player with the trial credits

Congrats on the launch! 🚀 What’s the most unexpected personal app someone has built with Glaze so far?

Creating apps I can use instantly from just a few sentences is genuinely fun and kind of unbelievable. I've already built two apps I actually needed. Congrats on the launch!

the site-specific-browser-meets-generative-AI framing is a great way to put it. curious where the generated apps actually run — fully local on the Mac, or phoning home? that boundary is the whole ballgame for anything touching personal data.

How does app signing and all work?

Already a user and lover of Raycast for years, Glaze is such a nice tool! Already built a personal finance tracker and many ideas to come 🤗

love that raycast is going after this, the "lives in your dock and works offline" part is the actual hard problem most AI app builders skip since they're all web output. does the generated app get real native access (menu bar, keyboard shortcuts, notifications) or is it more of a sandboxed webview wrapper under the hood?

I installed this app and created a menu bar app that displays the current weather conditions from my weather station in less than 15 minutes. It's great for those cases where you just want a simple local app on your machine and don;t want to fuss with more complicated tools.

Huge congrats for Shipping 👏 @frantzlight qq. are these conversion metrics available via the Loops API or they currently viewable only inside the web dashboard?

👀 Interesting. Planning to give this a try with my Hermes agent later today!

This is awesome. Love that once the data's available you can feed it to your agent (ideally through Basedash 😄) and figure out the best way to optimize.

The analytics breakdown by loop is really handy for spotting which campaigns actually drive revenue. Clean interface too, no clutter.

Finally tried Loops yesterday and the drag-and-drop email builder is genuinely faster than the clunky tools I've used before. Liked that analytics actually load without a delay.

Congratulations! Measuring whether a campaign actually drove the outcome is the part most email tools quietly skip, so it's good to see it built in. One question: for a longer activation window, can the attribution timeframe be set per goal, or is it a single global setting? Wishing the team a strong launch :)

Since Loops is built specifically for modern SaaS, I'm curious how Goals handles account-level vs. user-level tracking. If one user on a team clicks the email, but a different user on that same workspace completes the upgrade event, can the attribution logic tie that back to the original campaign?

Congrats on the launch! How does the attribution logic handle user-level vs. account-level conversions if a different teammate completes the upgrade?

The "opens and clicks are useful but they do not tell you whether a campaign worked" line is the frustration that killed my last three attempts at treating email as a serious channel. Every dashboard was clean and every meeting was hollow, because open rates went up but nothing downstream moved.

I run cold outreach rather than lifecycle email, but the same problem applies at a different layer. Reply rate is my proxy for outcome, but "reply" and "reply that led to a customer" are different metrics, and I have no good way to close that loop without stitching CRM data manually. Would be curious whether Goals is scoped only to product-usage conversion states inside Loops, or if there's a path to attribute against external systems (Stripe, Airtable, custom webhooks) for teams whose "conversion" doesn't live inside their SaaS.

Also the attribution window as a first-class input is smart. Most tools treat it as "check 30 days" without letting you match it to your actual sales cycle. Marketing-attribution problems mostly come from mismatched windows, not missing data.

I really like the slideshow-style website.

@frantzlight Chris, opens and clicks have always felt like comfort numbers to me, so getting a straight answer on whether a campaign actually moved people is genuinely refreshing to see.

Quick question do you support a holdout or control group, so you can tell the email actually caused the conversion vs. it would've happened anyway?

How does Loops handle the deliverability side of things compared to something like Postmark, especially for transactional sends that need to land fast?

How does Loops handle deliverability for transactional emails compared to something like Postmark, and is there a usage tier that makes sense for someone just starting out with maybe a few thousand sends a month?

This is a healthy move away from vanity email metrics. For small SaaS teams the useful question is not whether an email got opened, it is whether it helped the next operational state happen: activated, paid, rescued, renewed. The tricky part is showing enough attribution confidence that a founder can act without pretending the campaign was the only cause.

How does Loops handle deliverability compared to something like Sendgrid or Postmark, especially for transactional emails that need to land in the inbox fast?

opens and clicks are useful but they don't tell you whether a campaign worked is the most honest thing an email platform has ever said about its own metrics. being able to attach a real conversion goal to a campaign and see if it actually drove the outcome is what every email marketer has been doing manually in spreadsheets. having it built into the send flow instead of being an afterthought changes how you think about every email you send

How does Loops handle the difference between marketing and transactional emails under the hood, is it a single API with flags or two separate workflows?

How does Loops handle deliverability compared to something like SendGrid, and is there a free tier for smaller projects?

How does pricing compare to the big names like Mailchimp or Postmark once you start scaling up to higher volumes?

I like Goals connects campagins to real outcomes instead of only opens and clicks. how do you handle users who convert after using multiple channels? A clear attribution comparison could make the results even more useful.

Congrats on the launch! Love the shift beyond opens and clicks. How does it deal with overlapping campaigns where more than one email could have influenced the conversion?

So cool! I will try now :D

The "grows from local Claude Code token activity, nothing leaves the Mac" framing is what keeps this charming instead of turning into a productivity guilt-meter. Since there's no account or sync, what happens to a collection I've spent weeks on if I switch Macs or reinstall — is there a local backup/export, or does the gacha roster start over? I'd happily leave it running all day if the evolved species carry across machines.

The same should be done for people who code without AI, so they will be motivated to work on their own code more organically :)

launched my own product this week and spent the whole time staring at analytics dashboards, so a pet that just vibes next to the work instead of measuring me is weirdly exactly what I needed lol

the "heads home when it rains" detail is what sold me. that's not a feature, that's a personality

As a solo founder basically living in Claude Code all day, the local-only "nothing leaves your Mac" part is what sells me over another usage dashboard — a little companion growing alongside the grind beats staring at more numbers lol. Does it react differently to one long deep-work session vs a day of scattered short bursts?

Love that this one is proudly useless, that's the whole charm :) After a full day in Claude Code the last thing I want is another dashboard judging me, a little creature that just grows quietly alongside the work is a much nicer relationship with the tool. ;)

The detail I like most is the one that sounds like pure flavor: it wanders home to rest when it rains or night falls. Tying growth to token activity could so easily become a "grind more, feed the pet" guilt loop, a companion that also knows when to sleep says the exact opposite. That resting is what keeps it a friend instead of a productivity nag. Respect on the honesty too, being upfront that it reflects local coding activity and not your subscription % is a real trust signal.

Congrats on the launch! :)

the 'waves when a session's waiting on input' bit is the sharp part — in the local jsonl, blocked-on-a-permission-prompt and you-walked-away look identical. last event's an assistant turn either way, so the tell has to come from outside the transcript.

I think collecting twenty pieces sounds fun and rewarding. Can I choose one favorite pet that always stays active? a favorite option would help users build a stronger connection while still collecting every species.

Jason, this hits a very specific childhood nerve for me. The idea of a tiny creature quietly growing in the corner while I work made me smile more than I expected. Cozy little thing.

Heheh love this Jason! Guessing if you're thinking on monetizing or it's just for fun. Congrats on the launch anyway and wish you all the best!

The weather-reactive behavior is such a thoughtful touch, sending your pet home when it actually rains where you live makes the little guy feel alive instead of just decorative. Love that it all stays local too.

The no-account, no-upload framing is the detail that keeps this from becoming a productivity surveillance toy. For a local Mac companion tied to coding-agent work, I would keep the user-facing promise very explicit: what signal is read, where it is stored, and how someone can back up or reset it. That trust boundary is what makes small local tools feel safe to leave running all day.

The stressful world of builders could use more of this kind of fun.

The detail that got me isn't the gacha — it's that the pet perks up and waves when a Claude Code session is waiting on your input. I usually have two or three sessions running side by side, and "which one is blocked on me" is a problem I've half-solved with terminal bells. A pixel pet doing that job ambiently is honestly better UX.

Also appreciated the upfront note that it reads local coding activity, not subscription limits — that kind of honesty in a launch post is rare. Congrats on shipping, @besslframework 👌

how does it actually detect what im building with claude code, does it read the terminal output or something else

how does the gacha decide which species you get, is it truly random or does claude code usage somehow influence your pulls?

The real-time weather reactions are such a thoughtful touch — having your pet actually go home when it rains makes the whole thing feel alive instead of just decorative. Love that nothing leaves the Mac.

The little guy reacting to actual weather is such a nice touch, especially seeing him head home when it starts raining outside my window. Wish more desktop apps felt this cozy without asking for an account.

the gacha loop is genuinely charming, and i love that it actually reacts to the weather outside my window. caught it heading to bed when it started raining last night, which was a small but delightful touch.

Shipping export/import mid-launch because a commenter asked for it beats any launch video. Well done! I was wondering if token volume used to mean a human at the keyboard (but a growing share of Claude Code usage is agents grinding unattended overnight), then does an egg hatched by a background run feel earned?

the little guy actually heads home when it rains, which is way more charming than i expected for a desktop pet. curious to see what evolves after a week of coding sprints.

finally gave it a spin this morning and the weather reactivity is honestly charming, my little guy wandered off the screen right when it started raining. love that nothing leaves the mac

the weather reaction really got me, seeing my little guy head home when it started raining felt weirdly charming and the local-only thing is a nice bonus for something this cozy.

@besslframework This is super cute. Dolloproof just hatched. I get a feeling this is going to be fun
One question, does it track session limits or can it suggest optimal coding period, eg clubbing two sessions?
Also, one pain point on claude code is to remember putting a querry to start a session so i optimize the day better, can it do that for me?

I spend half my day inside Claude Code so this is dangerously relevant. Genuinely clever gamification - the pet growing with actual work beats every streak counter I've ignored. Does it react to failed runs too, or only progress? Might need to protect my pet from my debugging sessions. Congrats on the launch!

how does it actually detect what im building in claude code without uploading anything, is it just reading local session logs?

This is silly and not productive and I love it. Such a fun idea!

That's a cool idea!
But i can't see any close button to turn it off. Or exit from the app.

hey, product hunt! 👋

i'm enes, solo founder of osloq.

this came from a problem every developer knows too well. someone files a bug report, and before you can even think about fixing it, you drop what you're working on, dig through the repro steps, get your project into the exact state they described, and run it over and over just to confirm the bug is even real. half the time it ends in "works on my machine" and the issue sits there for weeks.

so i built osloq to do that part. you hand it a github issue, and it spins up a real sandbox, clones and runs your actual repo, and tries to reproduce the bug the way a developer would. then it hands you a report backed by real evidence. what it did, what happened, and whether the bug is actually real.

the hardest part while building it was trust. early on the agent would "reproduce" bugs that weren't real, basically confident hallucinations. so a lot of the work went into making it run real code and prove what it found, instead of guessing from reading the source. if it can't back a claim with evidence, it says so.

it's live today and i'd genuinely love your feedback. what would make this useful in your workflow? ask me anything, i'll be here all day.

Hey Enes, the reproducing part has always been the bit that quietly eats my afternoon, so seeing something take that off my plate genuinely made me pause. Feels like the kind of help I'd actually welcome.

The evidence-backed part is the important line here. For a small team, a bug-repro agent is useful only if the report can become a reviewable artifact: exact assumptions, commands, env gaps, and ideally a tiny failing test or script someone can rerun. That keeps it from turning into another confident AI opinion in the triage queue.

The "works on my machine" framing hits home — as a solo founder most of my bug reports come from non-technical users, so the "repro steps" are usually "it just broke" with no stack trace. Does Osloq still get somewhere useful when the issue text itself is vague, or does it lean on the reporter having written decent repro steps to work from?

Are there any plans to turn this into a full manual QA bot? Reproducing bugs are one thing but testing that the code is working as expected is also a tedious part I would love to do away with.

I like that the report is evidence-backed rather than just a polished summary. The detail I’d want in a team workflow is a clear “could not reproduce because…” state, so an issue can move back to the reporter without another manual triage loop.

How's it different than Linear's agent?

@enessss since Osloq stops at a confirmed repro with evidence rather than proposing a fix, is the output structured enough, or is it more of a human readable triage report at this stage?

the hard part with most repro bugs I've dealt with isn't running the code, it's the missing state - a specific user's data, a race condition that only shows up under load, third party API responses that vary. how does the sandbox handle issues that need real production-like data or timing to actually trigger, or is this mostly aimed at the straightforward "steps to reproduce" kind of bug

the setup loop is the sharp edge — the agent iterates until the repo's up. but a whole class of bugs is 'doesn't build on clean checkout,' and something whose job is to get it running will quietly patch past exactly that.

Running the actual code instead of reasoning about it is the right call. Curious how Osloq handles repos with external service dependencies - database calls, third-party APIs - where the bug only shows in a specific environment setup.

How does the sandbox handle repos with heavy infra dependencies like docker compose or database services. Does it spin those up too or just the app code in isolation?

I would like support for retrying failed reproductions. Multiple execution attempts could separate random failures from actual bugs.

Curious how it handles repos with external service dependencies like a database or third party API. Does the sandbox spin those up too, or do I need to provide credentials before it can run anything?

The "no evidence, no claim" stance is the right call — I build agents for commerce checkout flows, and verifying an agent's claims against what actually ran is where most of our engineering time goes too.

Two things I couldn't tell from the page: how do UI-only bugs work — if an issue only reproduces in a rendered front-end (dead button after hydration, checkout step breaking), does the sandbox drive a real browser, or is it limited to what's exercisable from CLI and tests?

And how do you keep the agent itself honest over time — do you re-run it against a benchmark of previously-verified issues when prompts or models change, so repro accuracy doesn't silently regress?

Congrats on the launch @enessss

Neat concept. Does it handle bugs that need specific user state or production data to actually reproduce?

Tried it on a tricky bug I'd been putting off and the agent actually cloned the repo and walked through reproduction steps I hadn't even considered. Loved that the report felt grounded in real output rather than speculation.

the no-setup sandbox thing is such a smart move, you skip the half-day yak-shave of replicating someone else's machine just to confirm a bug. nice execution.

Reproducing the issue is genuinely the worst half of the job - half the tickets I get are 'it doesn't work' with zero context, and you burn an hour just recreating the state before you can even start fixing. Automating the boring forensic part instead of the fix itself is a smart wedge. How does it handle issues that need specific data/state to trigger? Congrats on the launch.

Really interesting approach!!!!
Curious....how does Osloq handle issues that depend on external services or environment specific configurations?
Anyways...congrats on the launch! 🚀

The sandbox repro thing is genuinely cool, I picked a flaky issue from one of my repos and it came back with actual logs and the failing test. Wish I'd had this last week when I burned an afternoon chasing a "bug" that turned out to be a stale local DB.

How does it handle repos that need specific env vars or a database to actually run, does it ask me to configure those or just spin up a blank sandbox and fail?

Congratulations on the launch! The "reproduce before you claim" approach is a sharp answer to the trust problem with AI coding agents. When it reports that it couldn't reproduce an issue, how should a developer read that - "not a bug," or possibly a setup gap on the agent's side? 

the smartest thing here isn't the sandbox, it's what you DIDN'T build. every other AI dev tool promises to fix the bug, you scoped to proving it exists, which is the one step where "confident hallucination" gets caught instead of shipped

"if it can't back a claim with evidence, it says so" should honestly be the bar for the whole agent category

actually runs the repo in a sandbox to verify bugs, that part genuinely sold me. way more useful than the usual "here are 5 possible causes" guesswork from other tools.

Love the emphasis on "backed by real evidence" instead of confident hallucinations, that's the biggest friction point with most AI dev tools right now. The sandbox approach is solid.

Quick question though. How does it handle intermittent bugs or race conditions? Those are the worst to reproduce manually, but I imagine they're also tricky for an agent since a single run might not catch them. Does it retry, run multiple times, or is that something you need to flag upfront?

Really smart move focusing the agent on reproducing issues specifically, instead of trying to fix everything. Repro is the annoying part that eats up so much of my morning, and getting a clean repro is half the battle anyway.

Reproducing issues is genuinely one of the more annoying parts of the bug triage loop, especially when the report is half-baked and the reproducer assumes environment context the submitter forgot to mention. Curious how Osloq handles that case. When the issue is ambiguous or missing key details, does the agent ask clarifying questions, make assumptions and document them, or just fail gracefully and tell you what it couldn't figure out? Also wondering whether the reproduced case outputs something like a failing test or a script, or just a description of the steps it took.

The "spins up a real sandbox, clones and runs your actual repo" step is the hard part — most repros only fire with real env config, not just the issue text. To get a project runnable, does Osloq read a devcontainer/Dockerfile/build script or infer the setup itself? And for bugs that need secrets or a seeded DB, how is private-repo access scoped — a short-lived token per run, or standing access?

Hey everyone! 👋 I'm Salah, the maker of Archify.

I built Archify because I believe understanding software has become harder than writing it.

Instead of digging through DevTools, network requests, and source maps, Archify helps you see what's behind a web app its components, APIs, scripts, and architecture right from the browser. It runs entirely locally, so nothing leaves your machine.

I'd genuinely love your feedback, ideas, or even criticism. Thanks for checking it out! ❤️

does the extension need any permissions beyond the active tab, and can I audit that?

Salah, I'm endlessly nosy about how the sites I like are actually put together, so this scratches a real itch for me. The fact that it all stays on my own machine makes it even easier to enjoy poking around.

Love how clean the browser integration looks here, feels like the dev tools panel finally got the respect it deserves.

the way you can trace APIs right in the browser without leaving the page is genuinely clever, the inspection overlay feels really well thought out

How does Archify actually get visibility into the components and APIs inside a live app, does it require installing a script or does it hook into the network tab somehow?

Being able to peek into a running app's components right in the browser without setting up a separate debugger feels like a huge time-saver, especially for onboarding to a new codebase.

How does this actually work under the hood, does it inject something into the page or proxy requests through your servers? Trying to figure out if my proprietary code or sensitive data ever leaves the browser.

Love how Archify just lets me pop into a live app and trace the components without any setup. The browser-side inspection is genuinely smooth and feels like the dev tools experience devs actually want.

Finally a browser tool that actually helps me see what my React app is doing without constantly alt-tabbing to devtools. Loved clicking through a component tree and instantly spotting a stale closure bug.

How does it handle larger apps with heavy traffic in terms of performance impact, and is there a notable spike in load times when using it?

Being able to poke at components and APIs right in the browser without jumping through dev tools is genuinely useful, saves me a lot of tab switching.

The way Archify surfaces component behavior right in the browser is genuinely clever, saves a ton of tab-switching when debugging. Clean execution on something that could've easily turned into another cluttered dev tool.

How does it handle authentication when inspecting components behind a login, do I have to paste in cookies or does it manage that for me?

Archify is listed across Developer Tools, AI Workflow Automation, and AI Agents, which makes me wonder about the actual entry point. Is this mainly a browser-based helper for understanding web apps/docs, or does it connect to code repositories and engineering context too? The “understand software” line is broad in an interesting way, so a concrete example would help place it.

HI Mohd,

This is impressive. Good job. Is it possible to have the pop-up open in a side panel https://developer.chrome.com/docs/extensions/reference/api/sidePanel
So that i can keep it open as i visit different pages or sites ?

Does this work as a browser extension or a standalone app, and is there a free tier or is it subscription-only from the start?

Pretty handy being able to peek at component trees and API calls without leaving the tab, the in-browser view saved me from digging through source files. Wish the filtering was a bit faster on larger apps though.

Nice this is the one, help a lot in my works about research the market and technology <3

The best dev tools don't add another dashboard they remove one. Love the direction.

niceeee

@salahxd If Archify reads the live DOM and network activity locally, how does it identify components and API boundaries across different frontend frameworks? is detection framework-specific?

The context engine adapting when I tell it I am on holiday is the part that would actually make me switch — most task apps just keep nagging with the same list. Two day-one questions: when I brain-dump by voice and it extracts the wrong priority, do I correct that in text, and does the correction actually teach the context engine, or is it a one-off edit? And is the learned context tied to one device, or does it carry over if I capture on my phone and review on desktop?

me being skeptical of AI apps lately, I appreciate the clear focus here. Where does my task data live?

i wonder about the learning part. When nxt learns my personal context, does that happen from my task history alone, or can I tell it things directly, like my work hours or my energy patterns?

what happens when I disagree with the pick? Can I skip and get a second suggestion?

talking to a to-do list only works if it nails intent from messy speech — half the time i don't know what's next either. is the parsing on-device or server-side? and does it read urgency from how i say it, or just the words?

Me wondering whether it works offline because that would matter during travel. Offline support could make daily planning much smoother.

The "one task at a time" approach is such a thoughtful cut against the typical productivity app bloat. Feels like the team actually sat with the problem instead of just stacking features.

Honestly the one-task-at-a-time view is the part that got me, it stopped me from staring at a list and actually doing something. Also weirdly nice that it picked up priority just from me rambling about my day.

Nice approach. Does it connect to existing calendars or does it work as a standalone task manager?

How does nxt handle tasks when you mention something ambiguous like "next Tuesday" without specifying a time zone — does it learn mine or default to device settings?

the "one clear task, one reason why" approach is genuinely smart - too many task apps drown you in options when your brain just wants a single next move. love that it files everything automatically instead of making you play secretary.

How does nxt actually handle the "between the lines" part in practice - does it just run everything through an LLM each time, or is there some kind of learned model on your device that gets smarter as you use it more?

the "infers priorities" part is the make or break feature here. extracting tasks from a brain dump is mostly solved, but deciding what actually matters requires context the model doesn't have. like whether the thing due friday is more important than the thing my boss casually mentioned yesterday. how does nxt learn that? does it ask clarifying questions upfront or adjust based on what i actually complete vs snooze over time?

How does it actually decide what to surface as "the one clear task" when several things feel equally urgent?

This looks really interesting. How easy is it to correct when it gets a task or priority wrong? Good luck and congrats!

I use just Apple notes for my to do list but just speaking what I want to add when I want to add is a neat idea. Cool that it learns your habits!

I've abandoned probably 6 task managers at the exact stage you describe, organizing became the task. the "one task, one reason" framing is the first pitch in this category I've actually believed in a while. trying it today

I absolutely love the explicit 'ADHD friendly' tag on your launch assets. From a Go-To-Market perspective, did you find that leaning heavily into neurodivergent accessibility and minimizing cognitive overload helped you cut through the noise of the crowded productivity space faster?

The "one clear task, one reason why" framing is what actually caught me. For me the failure point in every task manager was never capture, it was the decision paralysis of staring at a 40-item list and abandoning it. Two genuine questions: when I brain-dump a rambling voice note with five half-formed thoughts in it, does nxt split that into separate tasks or treat one recording as one task? And how transparent is the "reason why" — is it surfacing the deadline/priority it inferred, or my stated context (heads-down week, travelling)? Keeping that recommendation from feeling like a black box seems like the whole ballgame. Congrats on the launch.

The "get what's next" framing is interesting because it implies the app is making prioritization decisions, not just surfacing tasks by due date. That's the actually hard part of any task system, knowing which of the 40 things on your list is the right one to start right now given context, energy, and dependencies. Curious whether nxt is doing genuine prioritization reasoning or whether "talk to your to do list" mostly means natural language input and query. Also wondering how it handles tasks without clear deadlines or urgency signals, since those tend to be the ones that rot at the bottom of every list forever.

This is promising. Curious if I can collaborate with my team or friends to create shared tasks or checklists. Congratulations on your launch.

Hey Product Hunt 👋 I'm the maker of Vox. I use GitHub Copilot constantly and got tired of being pinned to the keyboard, so I built a way to just talk to it. Run /vox and a reactive orb opens in its own window — you speak your turn, the session hears it, and the reply is read back. Voice in, voice out. You can barge in by voice to interrupt and correct it, there are live captions and a transcript, and it even reads your typed replies aloud. It works in the Copilot CLI and inside the Copilot app. It's pure JavaScript with no build step — it uses the browser's Web Speech APIs by launching Chromium in app mode instead of shipping Electron — so it installs in one line on Windows/macOS/Linux. Free and open source (MIT). I started it as an accessibility-minded experiment (a hands-free way to drive an agent), so I'd especially love feedback on the voice timing and the interrupt flow. Ask me anything!

Homepage: https://aasis21.github.io/vox/ · Code: https://github.com/aasis21/vox

Launching Chromium in app mode to borrow the Web Speech APIs instead of shipping Electron is a clever way to keep it pure-JS and one-line installable. The tradeoff I'd want to pin down: Chrome's SpeechRecognition streams audio to Google's servers rather than transcribing on-device, so for someone dictating code and context into Copilot, is any of the voice actually local, or does every utterance leave the machine? And do the transcript and captions get written to disk anywhere, or are they in-memory only per session?

voice in / voice out for coding is the interface i keep wanting — the friction was always latency and round-tripping audio to a server. are you running the speech piece locally or in the cloud? been deep in on-device voice on my side and the tradeoffs are brutal.

The reactive listening orb in its own dedicated window is a really nice touch, keeps the voice interaction feeling like a proper companion rather than just another terminal pane.

Voice for coding agents gets compelling when interruption and correction are first-class, not an afterthought. The agent is going to misunderstand file names, symbols, and intent sometimes; the useful workflow is being able to stop it, restate the constraint, and keep the same session alive without touching the keyboard. Nice to see barge-in called out explicitly.

That's clever. Any plans to support other AI coding assistants beyond GitHub Copilot?

Me appreciate the simple setup process. Why not include offline support? I think limited offline features would increase reliability.

Does the orb stay open in the background while I keep coding, or do I have to keep invoking /vox every time I want to switch from typing to talking?

How does it handle accents or noisy environments in practice, and is the voice model running locally or hitting an external API that could add latency or cost per conversation?

launching Chromium in app mode instead of shipping Electron is such a clean hack, one-line install with no build step because the browser already has the speech APIs. more tools should steal this

the barge-in interrupt is the detail that makes voice actually usable btw, nothing worse than waiting out a wrong answer

The voice input part is straightforward enough, but the interesting question is how well it handles the parts of coding where spoken intent gets ambiguous fast. Saying "refactor that function" out loud works fine when context is obvious, but what happens when Copilot needs clarification and the back-and-forth becomes a longer conversation? Curious whether Vox supports that kind of multi-turn dialogue or whether it's essentially one-shot voice-to-prompt with no correction loop. Also wondering how it handles things like variable names, file paths, or syntax that's painful to dictate accurately.

Hey Product Hunt 👋

I’m Daniel, and I’ve been building Notta for the past 6 years.

In that time, we’ve seen meeting transcription become part of everyday work. But we’ve also heard the same concern from many users: they want AI meeting notes, but not every conversation should be sent to the cloud.

That’s why we built Notta Desktop!

With Privacy Mode, Notta Desktop records and transcribes meetings locally on your computer.

🔒 No network required.

☁️ No cloud processing.

💻 Your audio, transcripts, notes, and related files stay on your device.

The biggest benefit is not just privacy. It is freedom from transcription limits.

Because transcription runs locally instead of through cloud infrastructure, Notta Desktop can support unlimited local transcription and note-taking. You can record long meetings, frequent calls, interviews, internal discussions, and sensitive conversations without worrying about cloud-minute limits.

Notta Desktop works for both in-person and online meetings. It can capture system audio directly, so you can record Zoom, Google Meet, Teams, Slack, Webex, and other calls without inviting a meeting bot.

🤖 No awkward bot in the room.

👥 No participant list clutter.

⚡ No extra setup for attendees.

Notta Desktop offers two local transcription options depending on your device and language needs: a Notta offline model for English, Japanese, Simplified Chinese, and Cantonese, and Apple Speech Analyzer on macOS 26+ for 11 supported languages: English, Japanese, Korean, German, French, Spanish, Italian, Portuguese, Cantonese, Simplified Chinese and Traditional Chinese.

🚀 Because everything is saved locally, your transcripts are also easy to use in follow-up workflows with tools like Codex or Claude Code: summaries, reports, PRDs, CRM notes, action items, or any custom workflow you want to build.

We made this for customer calls, legal reviews, financial discussions, internal strategy meetings, research interviews, and any conversation where privacy, control, and unlimited transcription matter.

Would love your feedback: where would offline, unlimited meeting transcription fit into your workflow?

Awesome solution! Could you please list the minimum specs required for the offline mode? I'm guessing you're launching a local MLX model?

@danielwayne This is interesting. I’ve always felt AI transcription tools are useful, but privacy is the part that makes me hesitate, especially for internal meetings or sensitive client calls.

A desktop-first, more private workflow makes a lot of sense. Curious how much can run locally today, and where the line is between local processing and cloud features.

Congrats on the launch. Will be following this one.

The unlimited local transcription is the part that genuinely matters to me — cloud-minute caps are exactly why I stop trusting a tool for long interviews and back-to-back calls, so removing that ceiling is a real unlock. Two specific questions beyond the local-vs-cloud split TJ already asked: does the offline model handle speaker diarization locally, or does that still fall back to cloud? And what can I export the local transcript and notes to — is Markdown one of the formats, so it drops cleanly into a notes vault? Good to see privacy shipped as the default instead of an upsell. Congrats on the launch.

Local-only transcription is the right call — 'not every conversation should hit the cloud' is exactly why I've avoided the hosted notetakers. Practically: is the speech model bundled and running fully on-device, so Privacy Mode still transcribes on a plane with wifi off, or is it local storage with processing that still reaches out? And if I later turn on a cloud feature like summaries, is that an explicit per-note opt-in, or does enabling it re-route everything through the cloud by default?

this is a real need, I know a few people in legal and healthcare who can't touch cloud meeting note tools no matter how good they are, compliance just kills it. running transcription fully local usually means a smaller/quantized model though, how's accuracy on multi-speaker calls with crosstalk compared to the cloud version, or is privacy mode meant more for single-speaker dictation type use

Great team building a product that turns AI meeting minutes into real insights and actions

Congratulations on the launch! The no-bot, on-device approach is genuinely useful for sensitive calls. In Privacy Mode, does the app make any network calls at all (for example: telemetry, analytics, licence checks) or is it completely zero outbound traffic while offline?

Great product!!

"no awkward bot in the room" is a real selling point, half the resistance to recording client calls is the third participant named Notta Bot making everyone perform

capturing system audio directly is the right call. checking this out for client calls where the conversation can't leave the machine

amazzzing product~

How does the face recognition handle group shots with people who haven't opted in, and is the AI matching done on your servers or locally on the photographer's machine?

This is an excellent solution to a big problem in any conferences… I have used it in recent React Nexus... It was amazing. All the best with it. 👍

Congrats on the launch!

the FTP upload straight to guest albums is a genuinely nice detail, most tools in this space make you export from lightroom or your camera app first which kills the "live" part of live events. at a big wedding with a few hundred guests, does the face recognition hold up with people in and out of shadow/backlight, or does accuracy drop enough that photographers still have to manually sort some of it

I would have really loved to have it when my daughters were babies 🤩

Love the intent here, and the "no ads on your baby" stance really lands. One honest question, since I'd want this to still be around when my kid's older: with a pay-once model and no subscription or ads, how do you plan to cover the storage and bandwidth for years of photos, video, and voice memos as they keep adding up? Curious how you're thinking about keeping it sustainable long-term.

Nifty idea. Privacy and security are so underrated for such precious memories. Any plans for a mobile app also?

Sushrut, the promise of one private place for all those little firsts really touches me. Knowing a child's photos won't be surrounded by ads is exactly the kind of care I'd want in something this personal.

how does the privacy actually work here - is the data stored locally on my device or on your servers, and can my partner access the same journal if we both want to add entries?

I’m really bad at keeping stuff like this organized for my kids, this is a neat idea! I was just like “when did my first start standing up on his own?”

Hi Product Hunt!

I built OneShots because creating app store screenshots always felt much harder than it should be.

For every launch or update, I had to open a design tool, adjust layouts for different devices, rewrite copy for multiple markets, export everything, check the dimensions, and then do it all again when one small line changed.

OneShots turns that workflow into something much simpler: upload your screenshots, choose a template, edit your copy and branding, then export store-ready assets in bulk.

A few things I focused on:

• Fast screenshot generation without starting from a blank canvas

• Templates made specifically for App Store and Google Play listings

• One-place editing for copy, branding, and visual style

• Multi-device exports without manually resizing every screen

• High-quality PNG/JPG exports packed and ready to upload

The goal is simple: help indie developers, small teams, and app marketers create polished store screenshots without spending hours in design tools.

Would love to hear your feedback and ideas for what we should improve next.

One of the small but painful parts of shipping apps is how much time gets lost on screenshots.
It sounds simple at first, but once you need different devices, multiple screen sizes, localized copy, consistent branding, and clean exports, it quickly becomes a full design task.
That’s the problem OneShots is trying to remove.
Instead of manually building every screenshot set, you can start from conversion-focused templates, customize the content, and export everything in the right formats for app stores.
What I like most about this direction is that it is not trying to replace designers. It is helping founders and teams move faster when they just need professional, store-ready screenshots without repeating the same work again and again.
Excited to keep improving the templates, export flow, and localization support based on feedback from the Product Hunt community

The 'change one headline and it syncs everywhere' bit is what actually kills the Figma-resize loop for me — more than the templates, that's the real day-one win. When I come back weeks later for a v2 update, does OneShots keep my branding and template as a reusable project, or am I re-uploading and re-styling from scratch each launch? And for the multi-language export, does it pull from my own translation file, or do I type each locale's copy in by hand?

Does it actually pull the screenshots from the App Store / Play Store connect directly, or do I need to upload them by hand every time I update a build?

Uploaded a handful of iOS screenshots and had localized Play Store images ready before my coffee cooled down. The bulk export is genuinely the part that saved me tonight.

That's a solid tool which will definitely help save a lot in the long run. Are you also planning a mobile app with push notifications for the alerts?

Finally a tracker that handles the random currencies I pay for without making me convert everything manually. The monthly true spend view is genuinely useful and refreshingly uncluttered.

Finally cancelled three subscriptions i forgot about after seeing them in the dashboard. Setup took about two minutes and the currency conversion just works.

does it auto-detect subscriptions from my bank or do i have to add them manually?

Does it pull subscriptions automatically from your bank or card accounts, or do you have to enter each one manually?

nice work. i wanted to build something like this. currently using notion database to track my work. i'll give it app a try.

finally something that doesnt make logging work feel like a chore. love that its just a text box, ended up using it for two days straight without thinking about it.

How does it handle data if I switch to a new phone, since everything is stored locally?

does the local storage mean everything stays on device or are widgets pulling data from a synced backup somewhere?

Kind of morbid but still a good idea to focus on what matters!

How does the widget handle the timeline when you add a moment in the past, does it shift the whole week grid or just slot it in without moving anything else around?

The widgets are genuinely the best part — seeing the weeks laid out on my Lock Screen makes time feel less abstract somehow. Appreciate that everything stays on device too, no account fuss.

How does the timeline update if you change your birthdate or want to add multiple life segments down the road, and is that adjustment something you can do from the widget itself or only inside the main app?

the local-only data approach for something this personal feels like the right call, especially paired with widgets that actually fit the home screen aesthetic instead of fighting it

Finally a way to pull exact numbers from NIFTY 50 filings without opening ten PDFs, and the source citations back it up nicely.

The custom connector setup is genuinely clever, makes financial research feel less like detective work. Love that it pulls directly from primary filings rather than scraping summaries.

Finally someone solved the PDF digging problem for Indian stock research. Asked it about a recent NIFTY 50 earnings call and it pulled the exact margin quote with the source page in seconds.

How recent are the filings it pulls from, and does it update in near real time when a new earnings call transcript drops, or is there a lag?

The Custom Connector setup being that quick is genuinely impressive, especially for something pulling from dense filings.

the fact that it actually ships a working backend with auth and a real database from one prompt is wild, most "vibe coding" tools still leave you wiring up Supabase at 2am

How does the AI agent handle complex business logic, like multi-step workflows or conditional branching, when you're not actively guiding it? Trying to understand how much hand-holding it really needs before you get something production-ready.

Congratulations on the second launch! Good to see code export and ownership also stated up front - that's the first thing I'd worry about. You list a wide range, from landing pages up to CRM and booking systems, so I'm curious where the practical ceiling sits. At what point does app complexity outrun the AI, so you're editing the exported code by hand rather than describing changes in plain English?

Ran my site through it and the citation readiness breakdown was surprisingly specific, pointing out schema gaps I didn't know existed. The copy-paste fixes saved me from digging through docs.

How does AISight actually detect the AI crawlers in practice, does it rely on server log access or just passive DNS sniffing, and does that work for sites behind a CDN?

love the UI!

Also the fact that it gave my website a 100/100 score pretty much across the board made me happy NGL

It did surface some fixes around trust headers which was a quick fix. Thanks

Does the executive report get generated on demand or do you run scheduled crawls, and how often does the underlying data refresh so the citation readiness scores actually reflect current site changes?

The schema fix suggestions were actually paste-ready and made sense, which surprised me after seeing so many SEO tools that just yell about missing tags. Clean executive report too, no fluff.

The chat-first approach is genuinely clever, especially for owners who'd never touch a traditional scheduler. Love that the approval step happens right in the conversation instead of bouncing to another screen.

A few things I didn't fit into the intro but You usually ask about:

There's a quick onboarding chat first, where you tell the assistant about your business - what you do, your tone of voice, who your customers are. That context then shapes every post, so it doesn't read like generic AI.
From there the social media assistant keeps learning your style as you go: edit manualy a draft or give feedback ("too formal", "write shorter on Facebook") and it picks up those preferences for next time.

It can also do its own research when a post calls for it - if the topic is more complex or needs current, real-world info, it looks things up instead of making it up.

And for anyone juggling more than one business: SyncBooster is built around brands and channels. A brand is a single business with its own knowledge base, brand voice and assistant, so multiple clients stay fully separate. A channel is a set of connected accounts within a brand - and one brand can have several, e.g. one publishing in Polish and another in English, each with its own Facebook, Instagram, LinkedIn and Google Business. So from one panel you can run many businesses, in many languages, without anything bleeding between them.

Does the AI keep learning my brand voice over time, or does each post start fresh from the photo and note I send?

The chat-first approach feels surprisingly natural. Sent a photo of a freshly painted storefront with a quick note and got a solid post in seconds, ready to publish with one tap.

How does it handle the brand voice over time once I publish a bunch of posts - does it actually learn my style or do I have to keep re-explaining it in the chat?

how does the pricing actually compare per minute to something like deepgram once you burn through that $5 credit

Latency on the TTS endpoint was noticeably snappier than what I was getting elsewhere, and the $5 free credit was enough to actually test a real use case. Solid option if you need voice inference without the usual hyperscaler pricing.

tested the stt endpoint on a noisy podcast clip and it came back clean way faster than i expected for the price. nice to see an api like this coming out of india.

The custom skill generation is the interesting part. Most code-quality tools produce a report the developer has to internalize and remember to apply next time. Piping the diagnostic back into the agent as project-specific skills is a more honest loop, the agent produced the fragility, the agent gets the fix.

Building MotionFy solo with Cursor, the pattern I keep hitting isn't obvious bugs (those get caught fast) but subtle drift, the codebase slowly starts violating conventions I established in month one because the agent doesn't remember them and I don't re-prompt them. That's the class of debt that gets expensive later, and it's exactly what a project-specific skill pack should catch.

Curious about the closed-loop metric side, when you rescan and detect "measurable improvement," what's the definition? Reduction in specific issue types, LOC of hotspot code, or something else? Trying to figure out if the skill packs actually train the agent or just prevent recurrence of the last problem.

seems like strong to delve deep into security. But the what’s the difference between a skill that checks security + fable 5 vs your app?

How does it actually detect the hidden stuff - is it running static analysis, LLM-based review, or some combo of both under the hood?

ran it against a side project and it flagged a sneaky sql injection i had missed in a rush. the skill pack idea is clever, feels like a real feedback loop instead of a one-off linter.