PH热榜 | 2026-08-23

Hey Product Hunt 👋 I'm Ankush, co-founder of Construct.

My last dev tool hit 30k users and got acquired. Writing the code was never the hard part.

Being the CRM, the support inbox, the follow-up guy and the fundraiser at the same time was.

Hiring was the obvious fix, but the runway math said no.

So when AI agents got good, I tried all of them. Every one reasoned for a minute to do a ten second task, burned tokens like they were free, and I spent more time fixing its work than doing my own. I had enough, so Nischal and I built the version we actually wanted.

Construct isn't a tool you open. He's a coworker who never clocks out.

→ He already has the tools your business runs on, and if he needs a new one he installs it like an app.

→ Once a job runs right he locks it in. Next time it's one command, no re-thinking, no token burning, and your team can trigger it too.

→ He remembers you, your business and everyone he talks to. It all lives in a secure enclave that only you and him have the keys to.

Brief him like a remote hire on Slack, close your laptop and he still keeps going.

For our PH supporters we are offering:

7 days free on Pro plan, then 40% off for the first year, or 20% off monthly if you want him to intern for the quarter.

The fastest way to see what Construct can do is to just try it.

Drop the thing that eats up your week in the comments and we’ll tell you how we’d automate it with Construct.

Want to see it working for your business?

Grab 15 mins and we’ll demo it live with you: https://cal.com/construct/15min

Or come hang with us and the team in Discord: https://discord.gg/puArEQHYN9

ps. much love to @fmerian for hunting us and to the PH team as well 🩵

Hey PH 👋 Nischal here, the other half of Construct.

We've been heads down on this since April, so really excited to finally have it out in the open.

I helped build the harness and the Product design. Before Construct I was working on decentralised hosting and compute solutions ,deep in isolates and WASM, which is where most of how Construct runs under the hood comes from.

In these months we've watched plenty of much bigger teams ship at the same problem, and honestly it's only made us more confident that our take on it is the right one. Being small means we get to make the weird calls.

Really excited to hear what you all think, and happy to answer anything today. Infra, cost, product, the parts still rough.

Wish you a successful launch! :)

It feels great to see a computer working for me directly...😀

Really cool! How will you compete with likes of Grok Bot and YC's QM?

Congrats on the launch! Curious how you think about something like Grok Bot, do you see it as a competitor to Construct, or are you solving a different layer of the agent stack?

One of a kind! Congrats on the launch 💯

@fmerian Thanks so much for hunting us, Flo ! 🙏 Really appreciate you putting Construct in front of the PH community.

We’ve been working on this since March, and it’s been exciting (and slightly terrifying 😂) to finally put it out there.

Excited to hear what everyone thinks and hopefully get some honest feedback today! 🚀

LFG!!! Congrats on the launch :)

Hey Product Hunt 👋

I’m Vibhansh Alok, the lead designer at Construct.

After working with agentic tools for almost a year, I’ve realized that many AI products eventually become just another wrapper around existing tools.

Construct was different from the start.

We didn’t just want to build an AI employee that takes prompts and executes tasks. Because then you end up constantly checking: Did it finish? Did it stop midway? Is it hallucinating?

At some point, being a developer started feeling like parenting your agents

What we wanted to build wasn’t just an employee you delegate tasks to. We wanted to build a 10x coworker that actually takes work off your plate. That’s why Construct is built around the principle of coworking

As a designer, I tend to look at tools from a slightly different perspective. I care not only about whether something works, but whether it actually feels relevant or even necessary in the way we work.

Humans are meant to create, think, and build.

We shouldn’t be spending our time digging through logs, managing emails, moving documents around, or maintaining repetitive workflows. You should be building and improving what matters, while your coworker is always there to handle the operational work in the background. That philosophy is also reflected in Construct’s design language, that is, intelligence should be something you harness, not something you constantly manage.

So, don’t just build another chatbot or a wrapper. If your tool has the potential to do more, let it. Scale it until it changes how people actually spend their day.

Would love to hear how you all perceive it. Cheers !!

Congrats on the launch, glad to see PostHog users building useful stuff!

Congrats on the launch! Giving AI agents dedicated workspace environments instead of clogging local systems is a smart approach to hands-off execution. Curious to see how it handles long-running, multi-step browser tasks over time.

Locking a job in once it runs right is the part I'd have led with. Re-planning a solved task on every run is where credits actually disappear, and most agent products quietly bill you twice for the same thinking. What I'd want to know is what happens when the locked path breaks because an API changed under it. Does he notice and re-plan, or run the stale steps and report success?

Hey, I have instead trying to test it but seems stuck in getting simple tasks like reframing a csv file to match a shopify template, etc

Raj here, one of the co-founders.

Writing the integration stopped being the hard part. Checking that it actually works is the whole job now, and that's the half your agent can't do.

The gap

Your agent can write a Stripe integration. It can't run one. It writes the code, tells you it's done, and you find out in production whether that was true. FetchSandbox gives the agent already in your editor two things it doesn't have: somewhere real to run integration code, and a way to prove the fix worked.

Why the proof check matters

A customer paid for 5 seats. A retry gave them 10, then 15. An agent fixed it, and after the fix nobody got any seats at all. Tests still passed because the duplicates were gone. Almost any fix makes the error disappear. Far fewer make the data right.

So the gate asserts the exact end state a correct implementation leaves, and refuses to go green when it can't reproduce the bug first.

Setup

One block in your MCP config. No API key, no signup. Works in Claude Code, Cursor, Cline, Windsurf, and Codex. 70+ ready-made sandboxes: Stripe, HubSpot, Clerk, Resend, Twilio and more, free to try.

We hit #3 on our first launch. The ask afterward was exactly this: don't just give me a sandbox, tell me my fix actually worked. This is that.

Has your agent ever confidently fixed something that was still broken?

"almost any fix makes the error disappear, far fewer make the data right" is the whole thing, and it is the same shape as the problem we keep running into.

to answer your question: yes, and the worst one was not even an agent. we added an anthropic key and three things were wrong at once. opus rejects an explicit temperature outright, one haiku model id had been retired and returned 404, and our own code sent a temperature on every call. nothing failed in testing because nothing in testing actually called it. any customer who had selected opus would have had every single reply fail on the first try. we were offering an integration nobody had ever executed.

the related one scares me more. we ran eight models against a live pricing api and two of them read the wrong row of a price ladder that was sitting in their context. one quoted 39.00 for an order that costs 9.60, the other quoted 9.00. the 39.00 gets caught by anyone glancing at it. the 9.00 does not, and that is the one that reaches a customer.

so the thing i would want to know about the proof step: does it assert the response shape, or the actual values? a 200 with a plausible wrong body is the failure that survives every check we have tried.

Testing webhook idempotency with AI agents is an absolute nightmare they always silently fail or fake the fix. Forcing the agent to prove it worked with an actual receipt URL before merging is brilliant. qq Are you planning to let us add custom internal enterprise APIs to the sandbox list soon? Upvoted...

Testing and verifying AI integration fixes in a sandbox before deploying saves so much headache. Congrats on shipping!

Best of luck in the launch day!

The end-state assertion is the strongest part here. In SaaS billing and automation work, preventing a duplicate event is only half the problem—the final subscription, entitlement, and audit state all need to agree. Does the sandbox also let teams test reordered events and delayed retries across the same workflow?

FetchSandbox addresses a real weakness in AI-generated integrations: verifying actual end states rather than merely checking for successful responses. The stateful sandboxes, failure injection, and proof receipts make this especially valuable for testing billing and authentication workflows before deployment.

One optional improvement would be showing a side-by-side state diff in each receipt, highlighting the expected and actual values across the full workflow. That would make subtle failures, such as a plausible but incorrect response body, easier to detect during review. Congrats on the launch!

what kind of integrations do you have?

Onboarding wrote a "you've seen the intro" flag one screen too early, so one click on Explore recorded the intro as done forever and every visit after that opened on step 2 of 2.

Fixed the write, then went through it on prod myself. On a fresh load the intro came back. The click wrote nothing, a reload got it again.

But everyone who had already clicked was still carrying the flag, and the rule reading it hadn't changed, so they stayed on step 2 permanently. Second deploy the same day to make the stale flag heal itself.

The fixture is the part I think about, set by hand to describe a returning visitor, so the buggy assumption went straight into the test and it passed right until the bug was fixed.

Does the proof step always start from a clean sandbox, or can you seed it with what the bug already wrote?

The 5 seats turning into 15 is the version of this that actually costs money. We hit the same shape at Zeplik, output that parses fine and passes every assertion we wrote, wrong in a way only the end state shows. Green CI on a bad write is worse than a red build because nobody goes back and looks. I'd want to know if the receipt catches partial failures too, 3 of 5 writes landing and the rest silently dropped.

Hi Product Hunt 👋

When we first launched Aximote here in May, it lived almost entirely inside the car. That was useful while driving, but it left one big gap: once you parked, there was no good place to understand what had actually happened.

Today, we’re closing that gap with Aximote for iPhone.

Every trip appears automatically in a personal dashboard. You can see what it consumed and cost, how efficiently you drove, how your results change over time, and how one drive compares with another — without an OBD dongle or manual trip logging.

For this release, we built:

📊 A customizable dashboard with more than 20 vehicle metrics
⚡ Our new Efficiency Score
🔌 Detailed trip, charging, consumption and cost analytics
📈 Trends and comparisons across trips and compatible vehicles
🚘 Support for multiple cars and one continuous driving history
🌗 A completely redesigned interface with light and dark mode and detailed 3D car models

The part I’m most excited about is not simply having Aximote on another screen. The iPhone app turns your driving history into something you can revisit, compare and keep across vehicles.

Your data lives with your Aximote account instead of being limited to a single car or manufacturer app. So when you manage multiple cars or change brands, you don’t have to start from zero again.

Since our first launch, more than 10,000 drivers have joined Aximote and together tracked over 1 million trips. Their feedback shaped almost every part of this release.

Download Aximote for iPhone

We’d love to learn from the Product Hunt community again:

After finishing a drive, what is the one thing you wish your car explained better?

— Laurenz, co-founder of Aximote

Actually, this app tells you a lot more about your driving than about your car. It’s a bit of a fitness app four driving skills - economic, sporty, efficient, long-haul, however you want it, both in the car and on the phone, across many cars and brands.

Congrats on launching today @laurenz_hinterholzer
Very interesting concept, how are your early users using Aximote? What are the most interesting insights they've gained?

Congrats on launching today @laurenz_hinterholzer

Very interesting concept what is the usp of Aximote compared to other OEM apps?

kudos @ shipping & launching 🎈

Having direct, mobile access to your car's telemetry data in a clean UI is super practical. Great launch!

no OBD dongle and works across brands is the part that stands out to me here - most of the trip logging apps I've tried either need a plug-in dongle or only work with one manufacturer's own API. is this pulling from each maker's official connected-car service under the hood (Tesla API, etc for each brand) with you handling the different formats, or is there some more universal telemetry standard newer cars expose that you're tapping into. asking because that answer basically decides which of my family's cars will actually show up in this and which won't

Congratulations on the launch! The app fills a gap many modern cars seems to fail at. With cars nowadays being computers on wheels with a ton of sensors, it's strange something like this has not been introduced earlier by car makers or similar.

I'm looking forward to follow your journey, good luck!

Hi everyone!

@aprilnea built something surprisingly awesome for Logitech hardware!

OpenLogi is an open-source Rust alternative to Options+ that talks directly to Logitech hardware over HID++ and UVC. It runs natively on macOS, Windows, and Linux, with the controls you’d expect from the official app and quite a few power-user extras.

The setup is nicely straightforward. Most things can be changed from the GUI, while the underlying config is just a plain config.toml file you can inspect, edit, or copy between machines. No account or telemetry required.

One small heads-up: quit Options+ before launching OpenLogi. Both apps need direct access to the same HID++ receiver, so running them together would naturally conflict.

Hey Product Hunt 👋 I'm AprilNEA, the person behind OpenLogi.

I bought an MX Master because the hardware is genuinely great. Then I installed Options+ — an account prompt, a background service, telemetry, several hundred MB — all so I could change what the thumb button does. My mouse settings shouldn't need a cloud round-trip.

So I wrote OpenLogi: a native Rust app that talks to Logitech devices directly over HID++. Remap every button (44 built-in actions), set DPI presets, toggle SmartShift, battery for every paired device. Bolt, Unifying, Lightspeed, Bluetooth or USB — macOS, Windows and Linux.

Two things I care about:

• Your config is a file you own. Everything lives in a plain config.toml. Read it, edit it, diff it, copy it between machines. No sync service, because there's no account to sync to.

• Nothing leaves your machine. No telemetry, and the update check is off by default.

What's honestly not there yet: per-app profiles (in progress), an Options+ importer, and Flow. If those are dealbreakers, it's not a full replacement yet — the roadmap is public.

If you use a Logitech mouse, tell me which model and what you'd bind to which button.

(Heads-up: quit Options+ / Solaar first — only one app can own a receiver at a time.)

Love that this is local-first with zero telemetry and no account, most peripheral software treats a mouse config as an excuse to phone home, so building it in Rust as open source flips that whole model on its head.

This is amazing, love finally having Linux support as well instead of having to boot to Mac to change configs.

Hey PH fam 👏

Excited to hunt Claude Academy by Anthropic today!

As models shift from simple chat to full agentic workflows, the hardest part isn't prompting - it's knowing how to build, delegate, and manage context. Claude Academy brings all of Anthropic’s official training into one free platform.

What stood out to me:

→ Agentic Architecture: Real guidance on Claude Code, subagents, and MCP

→ Role-Based Tracks: Specific paths for developers, creative pros, and leaders

→ 100% Free: Self-paced courses, cookbooks, and certifications for everyone

Check it out at academy.claude.com.

Question: Which topic are you most excited to master first- Claude Code, MCP, or multi-agent workflows? 👇

multi-agent workflows for me, no contest. I run Claude Code daily and the part I still fumble through by trial and error is knowing when to split work across subagents vs just doing it in one long context - I've under-split and wasted context, and over-split and spent more time coordinating handoffs than the task was worth. if the role-based track for developers has a real answer to that instead of just "here's how subagents work" that alone would be worth the sign-up. does it get into the judgment calls or mostly the mechanics of setting it up?

For a lot of beginners, Github can be quite complex and intimidating. And even for the advanced people, maneuvering through the Github Releases page to find the right installation file for their OS and architecture.

I've created Yatko for that reason. Yatko serves you the right installation file for your OS and architecuture. With Yatko, every installation is just one download button away. All you need to do is enter the name of the repo which you want to install on the frontpage, and Yatko will serve the download link to you directly.

That's not even the best feature of Yatko. Just change the domain name of any public Github repo to a Yatko, for example github.com/cli/cli to yatko.app/cli/cli and you'll be taken straight to a much simpler installation procedure.

Yatko is open source as well! So check it out and please let me know if I can make any improvements to it. Contributions are welcome as well!

Website: https://yatko.app

Repo link: https://github.com/argval/yatko (Make sure to star it if you like it!)

Tried the domain swap on two repos just now, yatko.app/cli/cli and yatko.app/argval/yatko, and both sit on "Loading release…" forever. Chrome 151 on macOS, and the network panel shows no request going out to api.github.com at all, so it's hanging before the fetch rather than getting rate limited. I'd look at the OS and arch detection path, that's the only thing running ahead of the call. Rough one to hit on launch day, flagging it early so you can catch it.

Best of luck in the launch day!

This is so smart! I'm a beginner myself, this thought once hit me but I never start to solve it because I thought maybe professional coders know how to download exactly and I'm just being stupid......I like this tool!

This fixes one of those tiny GitHub UX problems that almost everyone has run into at some point. The domain-swap idea is particularly clever :) How do you handle repos with unusual release naming or multiple valid binaries for the same OS/architecture?

love the idea of actually using the depth sensor for something other than diving logs nobody looks at twice. question though - I've got a regular Series 9, not an Ultra or Series 10+. is demo mode basically the whole experience for me (crown-driven water, no real depth data at all) or is there some lesser version that still reacts to being near water/humidity even without the proper sensor

Best of luck in the launch day!

Thank you,

We need more fun apps. I miss the Talking Moose, Oscar the Grouch, MacPuke, Racter... And what else was there for the old Mac? Now everything is so effin’ serious…

"Two weeks is a bold promise and I like it. As a solo app builder I respect the focus here, one skill done well instead of ten done halfway. Good luck with the launch!"

the "picture-story mnemonic per character" approach is the right call - rote repetition is exactly where I've bounced off other apps before. the part of kana that actually got me stuck on my own trip wasn't the characters in isolation, it was the lookalike pairs, shi/tsu, ki/ka, so/n, that I'd nail individually in an app and then blank on side by side on an actual sign. does KanaSensei's drill deliberately pit those confusable pairs against each other, or is it mostly one character at a time and the mixing happens naturally once you're further into the deck

What I look for most is friction at capture time - if opening a notepad takes more than the half-second it takes to open a new tab, I just won't use it, and the new-tab takeover approach solves exactly that problem.

One question: new-tab real estate tends to be contested. I've got other extensions that also want to override it (a dashboard, a speed-dial style start page). How does Tab Notes behave when something else is also fighting for the new-tab page - does it detect the conflict and warn you, or is it more "whichever extension loaded last wins" and you'd need to manually disable the other one?

Detecting upscales dressed up as FLAC is the feature I didn't know I needed, half my old library is probably lying to me. When it flags a suspicious transcode, does it just mark it or can it help me hunt down a proper replacement? Local music people are exactly the crowd that never wants a subscription, smart call.

Hi everyone! 👋

Local Music Organizer started as a small personal project for my own music collection.

I've been maintaining a local music library since the early iTunes years. When CD writers and MP3 car players became common, I converted much of my CD collection to 192 kbps MP3. At the time, that seemed perfectly reasonable.

Years later, after upgrading my audio equipment, I realized how much quality I had lost. I began replacing old files with better versions lossless releases, vinyl transfers, rare editions and music collected from many different sources.

Over the years the library became increasingly difficult to maintain. Tracks were replaced, moved and reorganized. Some files disappeared while old Music.app records remained. Playlists accumulated duplicate entries. Different versions of the same recordings lived side by side, along with inconsistent tags, missing artwork and metadata left behind by converters and download sources.

But Local Music Organizer actually began for a completely different reason.

I simply wanted a better playlist shuffler. I never liked how Music.app shuffle could return to the same artists too often while other tracks seemed to disappear for a long time. So I built my own system with spacing between artists, albums and related artist projects.

While building it, I started discovering all the other problems in my library.

That small playlist tool gradually became Local Music Organizer a toolkit for inspecting, organizing, analyzing, repairing and converting large local Music.app collections.

A major principle throughout development has been safety. Most analysis is read-only, and operations that can modify files require explicit user action and additional verification.

Everything runs locally on the Mac. There are no analytics, advertising or tracking, and your music library is not uploaded anywhere.

I'm an independent developer, and I originally built this because I needed it myself.

I'd especially love to hear from people who have also maintained an iTunes or Music.app library for many years: what problems have accumulated in your collection that existing tools still don't solve?

Thanks for taking a look!

The £220 to £520 line is the last thing in your description and it's the only part that pays for the app. A map of where I've been is a nice-to-have, money an airline owes me is a reason to install before I close this tab. I'd try flipping the App Store subtitle to lead with the compensation number and let the map be what people find afterwards. On-device with no account is the other thing worth saying louder, most flight trackers want my email before they show me anything.

I’m the indie developer behind ANCBuddy. I built this after repeatedly switching apps to control my QC Ultra on macOS. Happy to answer technical questions, share the roadmap, and collect feedback.

While I believe the main hook is the simple control of the headphones via Mac: I honestly like the AI Auto-EQ Feature a lot! It's so simple and really increases the quality and depth of music I listen to. It's unique (nothing similar seen so far) and makes a difference. Try it out!

Congrats on the launch. QC Ultra supports Bluetooth multipoint - paired to two devices at once and switching audio between them. Does the menu bar state (ANC mode, battery) track correctly when the active device flips, or is it scoped to whatever Mac you're on regardless of which device the headphones are actually talking to right now?

Hey everyone, Dmitri is here!

It feels so good to be back on Product Hunt after almost 10 years!

Today I’m launching Hyperfocus – a free, goals-first planner for Mac.

I’ve spent the last several years coaching focus to product people in tech, and I’ve always felt that task managers got productivity wrong.

Why? First, they're really good at capturing and organizing tasks, but they start from an inbox – teaching you to react instead of deciding what actually matters. Second, most planners don’t really have an opinion on how you should work, so many people end up with a long list of todos that makes them feel busy without making them productive.

Hyperfocus takes a different approach. First, it starts from goals – you decide what matters most in the next few months, then plan your weeks and days around them.

Second, it’s built on a simple belief: focusing on a few clear goals is better than checking off a long list of todos. So as you write, it gives you feedback on your goals and plans, and then helps to make them clear and realistic. 

So what’s inside?

• One workspace that connects long-term goals to weeks and days

• Grammarly-like feedback on your goals and plans

• AI for planning, prioritization, and goal setting

• Long-term, weekly, and daily planning rituals

• Focus sessions

• Local first, small, fast.

The app is free. Download it for Mac: https://hyperfocus.in

Hyperfocus is still early, and any feedback would be really helpful! If the problem and the promise resonate – I’d love for you to try Hyperfocus and tell me what you think. Does it fit into your workflow? What is missing? How does it make you feel?

That feedback would help a lot! Let's focus 🚀

The clean macOS-focused approach is refreshing. Having a planner built around doing fewer things well sounds useful. 👌

Congrats on the launch! Remind me a lot of Notion visually. Is there a reason why did you go with pages-like design instead of classic tasks lists?

Dima, congrats on the launch! Pretty cool to see that someone is optimizing for outcomes instead of just completing more tasks! Love the idea

Congrats @dmtsepelev ! the Agent is something that I missed in the previous version so now I am excited to get to my hyperfocus track - I need this much as never before

Congrats on the launch! This is one of those problems I’ve been trying to solve for years. Tried everything from Todoist/Things 3 to Sunsama and Claude/ChatGPT.

Been using Hyperfocus for a bit now and it looks really promising so far!

I like the approach - very similar that I use in my goals setup to d2d planning, but here it is implemented in a specific tool that guide you over all the process and control.

@dmtsepelev Love the goals-first approach! It’s so easy to get lost in endless to-do lists, so focusing on the few outcomes that actually matter feels refreshing.
Congrats on the launch!

Big congrats on the launch!

We’ve been thinking about a problem our sales team faces every day.

There are great tools for finding leads, another for contact data, another for email outreach, another for managing inboxes, and then a CRM to keep everything together.

But why does a small sales team need 5 to 6 different tools just to run outbound?

We’ve started working on an idea: one AI-powered sales platform that can handle the workflow from finding the right prospect → getting verified contact information → launching outreach → managing replies → tracking everything in CRM.

The bigger idea is to eventually have an AI sales copilot where you can simply tell it:

"Find me companies that match X, identify the right people, and help me start conversations with them."

Still very early and we have a lot to figure out.

Curious..... if you could automate one part of your sales process today, what would it be?

We started Oppora.ai for a simple reason: our own sales team was drowning in manual work.

List building. Finding contacts. Verifying emails. Switching inboxes. Running campaigns. Updating CRM.

Too many tools. Too much copy-paste. Not enough actual selling.

So we started building Oppora internally — one platform to simplify the outbound sales workflow.

After seeing our team use it in real-world outreach and generate opportunities, we realized this wasn't just our problem. Sales teams everywhere are dealing with the same fragmented stack.

So we built Oppora.ai 🚀

Instead of stitching together multiple tools, Oppora brings together:

🔎 Lead Search & Contact Finder — find the right prospects and contact information

🧠 Lead Intelligence — enrich and qualify prospects

✅ Email Verification — reduce bounces and keep lists clean

📨 Email Campaigns + Multi-Inbox — launch and manage outreach from one place

📊 AI-native CRM — manage prospects, conversations and opportunities

🔗 Integrations — connect Oppora with your existing sales stack

And the part we're most excited about: ORA 🤖

ORA is our AI Sales Copilot.

You can ask ORA to find leads, build an outreach campaign, help manage replies and move prospects through the sales workflow.

The bigger vision is simple:

Turn a sales goal into qualified conversations and booked meetings... with AI doing the heavy lifting and humans staying in control.

Oppora started as something we built for our own team. We then opened it to the world, learned from real users, and kept improving it.

Today, we're officially launching Oppora.ai on Product Hunt. 🚀

We're excited to introduce it to the PH community and would love your feedback — what you like, what you don't, and what you think we should build next.

Let's build the future of AI-powered sales together. 🙌

What stands out to me most about Oppora is the economics of the product.

A typical outbound stack means paying separately for a lead database, contact enrichment, email verification, campaign sending, mailbox warm-up, LinkedIn automation, reply management, and a CRM.

Each subscription may look manageable on its own, but the total monthly cost becomes heavy very quickly, especially when pricing increases with users, contacts, mailboxes, or credits.

Oppora brings all of these functions into one connected system. That makes it one of the most affordable ways for sales and marketing teams to run the complete outbound cycle.

Beyond the savings, it also creates a much cleaner tech stack. There are fewer subscriptions, fewer integrations to maintain, and less data moving between disconnected tools.

Head of Engineering at Oppora here, adding the build side. 👋

The thing that took longest wasn't the AI. It was making the agents work as one system instead of eight separate features.

One workflow, not eight buttons. Most tools give you AI, but you still trigger each step yourself. In Oppora you drag the agents into a workflow once and it runs on a schedule. Find companies, get decision-makers, verify emails, score them against your ICP, send, handle replies, sync to CRM. Email and LinkedIn run in the same sequence, so touches coordinate instead of colliding. No logging in every morning.

What makes that survive unattended is that agents pass structured data to each other, not text. Lead Finder receives a proper company object from Company Finder, so nothing gets re-interpreted along the way. Every step has retries and a defined failure mode, because a workflow that silently half-finishes at 2am is worse than one that stops and tells you.

Deliverability was the harder problem. Mailbox rotation, warmup, and verification before anything sends: every email is verified before it goes out, no guessed patterns, no unverified sends. Get this wrong and it doesn't degrade slowly, it burns your domain. Most of our infrastructure work went there.

Two things not mentioned above. Everything the workflows do, you can also just ask: Ora, our in-app agent, sits on ~125 tools across the platform, so "find 50 fintech CTOs in Berlin and add them to my Q4 list" is a sentence, not a process. And we shipped a Claude MCP server + REST API on the same key ⚡ so you can run lead search and outreach from Claude, Cursor, or n8n instead of our dashboard.

One question I'd like the PH view on: which parts of outbound should be fully automated, and where do you still want to approve before it sends? Our take is sourcing hands-off, first reply in draft mode. Curious if that matches how you'd actually run it.

I'm not on the sales side of this — I'm the project manager, and I run QA. So I came to Oppora from a completely different angle than most people here: not "will this send good emails," but "does this system actually hold together?"

I'll leave the outbound story to the sales folks. What I can speak to is what a stack of disconnected tools looks like from the inside, because testing one is a nightmare.

When your process is spread across five separate products — one for data, one to verify, one to send, one to warm up, a CRM to catch replies — you don't have a system. You have five systems and a pile of glue holding them together. And every seam between them is a place where things quietly go wrong. Data gets dropped in a handoff. Counts don't reconcile across tools. A field that's filled in over here shows up blank over there. Nobody owns the whole picture, so when something breaks, everyone points at the tool on the other side of the integration.

That's the part I care about. From a QA seat, most of the real defects in outbound don't live inside any one tool — they live in the gaps between them. You can't test a workflow that spans five vendors that don't talk to each other. You just hope it lines up.

The reason one connected system matters isn't that it's tidier. It's that there are far fewer places for the data to lie to you. When find → enrich → send → reply → CRM all run on the same rails, the numbers mean the same thing at every step, and there's one place to look when they don't. That's what makes it trustworthy enough to actually hand work off to a teammate, or eventually to the AI.

My whole job is to be the person trying to break this before you can. So take it from the resident skeptic: the thing worth caring about here isn't any single feature. It's that one system you can actually verify beats five you can only hope about.

If you're running a stitched-together stack right now, I'd genuinely like to know where it bites you — my money's on the handoffs.

👋 Hey PH! Head of Product here.

Sanjay covered the "what," so here’s the cheat code for getting the most out of Oppora on Day 1:

Don’t overthink your first campaign. Try this exact prompt in WORKFLOW right now:

"Find 30 marketing directors, verify their emails, and send a short, 2-step sequence asking for a 15-min chat."

Watch it search, enrich, send, and route replies, all without you touching a single tab. That’s the "aha!" moment we obsessed over building.

We’re hanging out in the comments all day to answer questions, squash bugs, and hear your feedback. So please: ask tough questions, request features, or even roast our UI. We build fast, and your feedback directly shapes what we do next.

Curious: What’s the most annoying, repetitive task you’re hoping to automate this week? Let’s hear it! 👇

Congrats on the launch! Bringing lead discovery, enrichment, verification, outreach, and CRM into one workflow could save small sales teams a lot of time and tool-switching. Wishing the Oppora team a great launch! 🚀

I am all open for feedback and suggestions, try it out and drop suggestions !

Great project man! I joined the board ❤️

Hi Product Hunt 👋

I'm a power user of YouTube — and honestly, why wouldn't you be? YouTube is a genuinely competitive marketplace. Creators get paid by how many people watch, so they're incentivised to do the research properly, check it, and present it well. Better work earns more views. Whatever you think of the feed, the good material on YouTube is very good, and a lot of it exists nowhere else.

So I use it deliberately. I go through my feed daily and drop anything worth keeping into a topic playlist — for when I actually sit down to go deep on that subject. When I find someone consistently doing real work, I subscribe, so more of it reaches me. That's 200+ channels and 50+ playlists now. YouTube caps a playlist at 5,000 videos, which I know because I've hit it. That's a knowledge base, and I built it by hand, over years.

And then I go to use it, and there's nothing there. YouTube gives me no tool that touches any of that effort. I can't search across my own subscriptions and playlists — every search is a global search of all of YouTube, ranked by parameters I don't control and can't switch off. Its AI will discuss the one video I happen to have open, and nothing about the thousands I chose. The library is mine. The one thing I can't do with it is use it.

That's what asktube is for. Ask across every channel you follow and get one answer with the videos it came from. Ask a single creator. Compare a few. Or ask it to build the playlist that takes you from beginner to competent on something — and then you go and watch it. asktube works out what deserves your time; it isn't trying to replace the watching.

The part I'd most like you to poke at: it's free, including the AI. Not trial-free — free, because the AI is yours. asktube runs on Bodhi App, a local AI gateway I also built: you connect your own model, approve exactly what asktube may use, and your keys never leave your machine. asktube is the proof that a website can use your local AI safely. Everyone else in this category rents you someone else's model for $8–20 a month.

There's a live demo — no account, nothing to install. Being straight about it: the channels in the demo are hand-picked by me, not yours, and the AI is capped by the hour and by depth, because I'm paying for it. Everything it does, it does for real.

What's rough: first answers take ~20 seconds, because it's reading rather than recalling. Mobile is newer than desktop. And the demo collection is a slice, not the product.

What's next: better cross-channel comparison, and the developer surface — asktube speaks MCP, so your coding agent can research your YouTube library directly.

I'd genuinely rather have your criticism than your upvote. Try the demo and tell me where it falls over — I'll be in the comments all day.

— Amir

Really helpful for my use cases. Easy to integrate with other tools.

looks to be a neat solution for missing content in ever growing youtube subscription feed
good luck

asktube is really helpful.

All the best for the launch Amir!

good luck 🚀 curious what happens when two people keep outbidding each other for #1

Awesome project, best of luck!

that honestly huge

3 founders already claimed their spots 👀

Who’s going to take the 4th spot? 👀🔥

The race is getting interesting. Who’s making the next move?