Hey PH 👋

I'm Jaspar, co-founder and CEO of Artisan. I started my first business at 7, selling candy from my bedroom. Since then I've raised $36M+ to build AI employees and we're currently at ~$10M ARR.

Some of you might remember our first PH launch in February 2024 when we first launched Ava the AI BDR.

Today we're releasing Ava 2.0. We rebuilt her from the ground up and she is now fully autonomous and she runs your entire outbound on autopilot. She finds leads, sends personalized outreach, handles objections, and books meetings on your calendar. No human in the loop.

She searches through 300M+ contacts, enriches every lead, and watches for intent signals like funding rounds and leadership changes. She launches personalized multi-channel campaigns, runs continuous multi-variate tests, and auto-optimizes toward what converts. When a prospect replies, Ava reads the response, handles objections, answers questions, and books the meeting.

With Artisan, the entire sales stack lives in one place. Lead discovery, enrichment, signals, sequencing, a dialer, and deliverability infrastructure. All in one platform.

She's already running outbound for thousands of reps at companies like Corgi, SaaStr, Quora, and CookUnity.

🎁 For the PH community: new users get $300 in free credits. No credit card required. artisan.co

Jaspar

with ava running fully autonomous multi channel campaigns on autopilot, how are you guys managing local domain safety under the hood? sending huge volumes of automated personalizations can flag spam filters fast if the inbox infrastructure isn't airtight. good job team

The hard part with an autonomous BDR is not “can it send more?” It is whether the system knows when not to send.

For a tool like Ava, I’d want the review layer to make a few things very visible before outreach goes out: why this account, why now, what evidence supports the angle, what would make this a bad fit, and what the fallback is if confidence is low. Learning from ignores and bad-fit replies matters as much as learning from booked meetings.

That’s the line between useful outbound automation and just scaling the same spray-and-pray problem with better copy.

Looks interesting. The data with Eva covers for niche industries like local small businesses as well or its more relevant for B2B startups and the likes??

How do you keep it from burning your domain reputation or sending the same outbound style to everyone?

Congrats. Since Feb 2024, what has changed and what is upcoming on the roadmap?

Over the last 3 months, I’ve tested 8–9 similar services. All of them had the same problem — the credits disappear within a few days, and the cost per contact ends up being quite high. Have you calculated the average cost per contact or other metrics?

fully autonomous outbound with no human in the loop is bold. the personalization has to be genuinely good or you're just automating spam at scale. curious what the reply rates look like vs a human SDR running the same list

Congrats on the launch @jasparcj @lucas_moeller ! Very cool, and timely! Upvoted :)

I have been juggling around with Apollo/Prospeo + Clay + ChatGPT and that's really messy!

When you say searches contacts and sends outreach messages - which social media handles are included? And how do you measure the output, do you have any numbers around how many of those outreach messages convert? And how it is compared to the traditional manual ways? Thanks!

The autonomy is impressive. The open question for me is the one Anna raised: what is the failure mode when Ava is confident and wrong? Outbound is unforgiving, because a confidently wrong message still gets sent and the cost lands on your domain reputation, not on a quiet retry. Curious whether 2.0 has a notion of "not sure enough to send," or whether everything above a threshold just ships.

I'm just starting out, I have a quick question: in the connected Gmail account I have multiple accounts set up - I was never asked which account will be used to send emails. This means either the default account will be used, or the main account will be used - but which is it?

Curious how you think about the economics of outbound if agents become abundant. Historically, the bottleneck wasn’t sending messages, it was hiring, training, and managing SDRs. If every company suddenly has an unlimited number of AI SDRs, does outbound become more effective? or does attention become the new scarce resource?

In other words, does the winner become the company with the best AI, or the prospect with the best filters?

Is this only for B2B ? Also, whats the source of the 300M+ contacts ? Is it across the globe or specific to geography ?

Congrats on the launch! Going from AI-assisted outbound to fully autonomous outbound is a pretty big leap. Curious to see how Ava handles the messy real-world objections that make human reps earn their coffee.

Super interesting, the enrichment is generated in app, or through APIs/?

Is it only for B2B?

Congrats on the launch! A few questions:
- How does she learn about my ICP, the different segments, and their specificities (buying journey, pain, etc.)
- Is Ava improving over time?
- What kind of outreach volume does she handle (via email & LinkedIn)? How does she ensure great deliverability?
- Is phone call on the roadmap?
- Among the 300M+ contacts database, what's most represented? (B2B? US?, etc.)

is it not working on mobile ?

Wanted to try out, still wasn’t able to pass onboarding page. After feeling all info, didn’t see next step (on mobile)

Been waiting for something like this. Setting up custom polling logic to watch external data sources is one of those tasks that eats 2-3 hours and nobody talks about

Been doing the "scrape on a cron + diff manually" thing for competitor tracking for months. The webhook approach is cleaner. One question/ does it handle pages that hydrate content via JS after first render?

Web change monitoring via webhooks is something I've wanted for competitor tracking for a while. Does it handle JS-rendered pages or only static HTML? A lot of the pages worth monitoring hydrate content after first load.

The compliance monitoring use case feels underexplored here. Regulatory pages, terms of service, and policy documents change infrequently but consequentially. The kind of thing you want an agent to flag immediately when it shifts, not catch on the next scheduled crawl. The challenge is those pages often have boilerplate that changes (cookie banners, footer dates) without the substantive content changing. Curious whether /monitor lets you scope the watch to a specific element or section of a page, rather than monitoring the full document. that would make it significantly more useful for policy/legal tracking workflows.

1. How does the tool bypass advanced anti-scraping blocks like Cloudflare or CAPTCHAs?

2. Can the system detect and highlight a single word change within a text?

@ericciarla @caleb_peffer @nickscamara Congrats on launching /monitoring!

The shift from "blind polling" to "intelligent diffing" is exactly what dev teams need to keep AI agent costs under control. I’ve lost count of how many tokens I’ve burned re-scraping unchanged docs.

How does /monitor handle dynamic content (like JS-heavy dashboards) vs. static text? Does it ignore irrelevant UI changes (like ad rotations or timestamps) to ensure the webhook only fires for meaningful data shifts?

🕸️

This is one of those features that seems obvious only after someone builds it. Monitoring changes instead of constantly re-scraping pages feels like a much smarter approach. Congrats on the launch!

Can you set rules for what counts as a meaningful change, like only when certain sections or selectors update?

agents re scraping entire pages every hour just to check if one price changed is such a waste of tokens and money. only ingesting the diff is how it should've always worked

Congrats Eric, Caleb, and Nick! 🔥

Love this build. The 90% token cut is wild. That's the number every team running AI products is quietly desperate for.

We pull buyer-side data into FireCoach all day long for sales prospect profiles, and the re-scrape vs. monitor tradeoff is exactly the thing we keep fighting. Going to share this with our engineering team.

Quick question: does the diff get summarized in natural language before hitting the webhook, or does the receiving agent still have to interpret the raw diff?

Either way, this is one of those "obvious in hindsight" products. Nice work!

Hey Product Hunt, I'm Tong, founder of Ava Studio 👋

If you’ve ever launched a product, you probably know the feeling: the product is live, the site looks good, users are waiting somewhere… and now you need ads.

That’s where a lot of founders get stuck.

You can stare at ChatGPT trying to write hooks. You can brief an agency that costs more than your MRR. You can spend three hours in Runway or CapCut and end up with one clip you’re not sure is even usable.

The problem isn’t that founders don’t know their product. It’s that making good ads requires an entire creative workflow: research, briefs, scripts, hooks, actors, product shots, editing, resizing, testing, and iteration.

We built Ava Studio to make that workflow feel like having a creative team on demand.

Drop in your product link and Ava studies your product, your market, and the ads already working in your category. Then she generates creative briefs, scripts, and 50+ editable ad variants ready for TikTok, Instagram, Meta, and YouTube.

What used to take weeks of creative briefs, scripts, edits, and agency back-and-forth now starts from one product link.

Watch me walk through it personally here:

https://www.youtube.com/watch?v=2RS2-3qaEhg

A few things people are surprised by when they first try it:

🔹 Ava starts where good agencies start: research. She studies your product, your category, your competitors, and the ads already winning before writing a single script.

🔹 It does not just make one video. One product link becomes a complete campaign: creative briefs, hooks, scripts, storyboards, actors, product shots, and 50+ editable variants.

🔹 Every output is remixable. Change the hook, swap the actor, rewrite the script, replace the product shot, or turn one good idea into ten new angles.

🔹 The template library is built around ads that already work. Browse winning formats by category, clone the structure, and customize it for your product in minutes.

🔹 Ava keeps the whole creative process in one workspace: research, strategy, scripting, generation, editing, and export for TikTok, Instagram, Meta, and YouTube.

Ava is built for founders and small teams who need ads that look good, move fast, and don’t require hiring an agency.

Every new account gets 500 free credits, enough to run a full workflow from research to finished variants.

Try it here: avastudio.com

Please break it, tell us what’s missing, and send us the weirdest product you can throw at it. We’re building this with founders who actually need to ship.

Much love,

Tong

being able to turn one winning creative angle into ten new variants with quick hook swaps is exactly what performance marketing needs right now. ad fatigue happens so fast on tiktok, so keeping the asset pipeline fresh is everything. awesome job buddy 👍

Congratulations. I actually made my first AI video using your product this morning. Best of luck to you all.

so all this means we don't need to hire a whole marketing team once I've use this app, right?

The research-first part is the right instinct. The risk with AI ad generation is not usually producing enough variants; it is producing 50 variants that all express the same weak claim with different visuals.

One thing I’d love to see in a workflow like this is an angle map before generation: audience, pain/risk, proof asset, objection, hook family, and why this angle is worth testing. Then when an ad works or fails, the learning attaches to the angle, not just the finished video.

That would make Ava Studio feel less like a content machine and more like a creative testing system.

Big congrats on the launch. The hardest part of short-form ads is usually coming up with enough strong creative angles consistently, so turning research and hook generation into a faster workflow is genuinely valuable for teams shipping large content. Good job.

Tong, you nailed the post-launch panic — "the site looks good and now you need ads" is exactly the wall most founders hit. Having something research the product, write the hooks, and hand back 50+ editable variants for TikTok/Reels/Meta beats burning three hours in CapCut on one clip you're not even sure about. Congrats on the launch! I'm @JayTheSong on X — would love to connect.

AI video ads for solo operators is the right use case hiring a creative team isn't an option but you still need content that converts. Does it handle vertical format for Reels and TikTok or mostly landscape output?

This is the part that feels most useful to me: starting with research, then turning it into multiple editable ad angles instead of just one polished clip. Very founder-friendly for testing creative quickly!

Congrats on the launch! We do both human and ai ugc as well. I'm curious who are the best customers you've had using Ava Studio and what are the sample campaign metrics, CPM and concrete accounts look like as lighthouse case studies?

Ok, the "stare at ChatGPT trying to write hooks" line made me laugh because I've done it this week. Multiple times.

The creative workflow tax on founders is one of those quiet killers. You think you're going to spend an hour on ads, four hours later you've got one mediocre clip and a headache.

Product link to 50 variants is the kind of thing that would have saved me serious time over the last few months. Going to share with our team.

Quick question: how does Ava handle brand voice? Does she pick it up from the site copy, or do you train her on a separate input?

Congrats on the launch!

This is very cool! Can I use this to generate talking videos or only non talking videos? I've been looking for studio to mass produce UGC content for our product and they often charge a lot. This solves a huge problem for us!

Love this, running out of ideas it’s so horrible and this just takes care of that.

It would be really cool to add competitor analysis and a database of best practices for each industry/product category. Because what matters is not just creating a beautiful video, but one that actually drives sales ;)

Looks to work well for physical products - how does it work for a B2C saas product?

@tpow Congrats on the launch!

How does Ava Studio ensure the generated hooks and scripts actually capture the unique voice of a specific brand, rather than just recycling trending formats?

Also, I'm curious if the "50+ variants" are truly distinct in narrative structure, or mostly visual/resizing iterations?

Congrats on the launch! Solving the creative angles bottleneck with faster research and hook generation is genuinely valuable. Well done.

Congrats on the launch, Tong and team! Love that Ava starts with research instead of jumping straight into generation. That foundation can make a huge difference when testing creative angles.

the post-launch panic of 'okay now we need ads' is way too real. 50+ variants from one product link means you can actually test angles instead of going all in on one creative and hoping it works

Hey Hunters 👋

I am excited to hunt Ahrefs Agent A today.

This feels less like a simple AI SEO tool and more like having a marketing teammate that actually understands search data.

Agent A can handle keyword research, competitor analysis, technical SEO audits, content planning, backlink insights, and even AI visibility tracking — all powered by Ahrefs’ huge search database.

What I really liked is that it focuses on execution, not just generating content. The integrations with tools like WordPress, Notion, Slack, and HubSpot also make the workflow much smoother.

Big congrats to the Ahrefs team on the launch 🚀

The useful leap here is not just “AI can execute SEO tasks,” it is whether it can explain priority and uncertainty well enough that a marketer trusts the next action.

For content planning especially, I’d want Agent A to separate a few things that often get blended together: keyword opportunity, SERP constraint, customer language, competitor gap, and the brand’s actual point of view. Those are different inputs. If the agent can show which one is driving a recommendation — and when it decides not to create content — that would be a big step up from dashboards or generic AI briefs.

Ahrefs data powering a marketing agent is a strong combo the data has always been the best part, the UI just gets in the way. Does it handle content gap analysis and suggest specific articles to write, or mostly technical SEO tasks?

We've spent a surprising amount of time manually watching HN whenever we launch something.

What signal ended up being most predictive for you — velocity of upvotes, comment activity, front-page movement, or something else entirely?

Congrats on the launch! Ahrefs has always had the data — adding an agent layer on top of it feels like a natural next step. Curious to see how much manual SEO work it can actually eliminate.

ahrefs data is already the best in the game so an agent that can actually act on those insights instead of just showing dashboards is a big step. the gap between 'here's what you should do' and 'done' is where most seo tools stop

As a formerish SEO (well, still an SEO inside, my scope has just expanded beyond organic, too), super excited for this launch. Ahrefs has been my go-to tool for years now but it's a hard one to learn + a pricy one. If Agent A can do what it promises, it would help bridge that gap of "hey, this tool is really really useful.... but it's also super hard to learn and turning the insights into actual helpful work is hard unless you have someone on your team with Ahrefs experience"

Hey Product Hunt 👋

I'm Midori, founder of FireCoach.ai. I've been in sales long enough that I've watched the same thing break revenue at every company I've touched. Two exits in, I finally stopped accepting it.

Here's the loop. You hire a rep. You onboard them, pair them with a manager, hand them the playbook (which they skim once and never open again, and you know this). The manager coaches the two or three reps they actually have bandwidth for. Everyone else wings it.

The ones who need help most are the last ones to get it. Usually caught in a deal review, after the pipeline damage is done.

Your best methodology lives in your top coach's head. There's only one of them. And they're already in 12 other meetings.

I talked to 100+ sales leaders in the last year. Every single one said some version of "I can't be everywhere at once." A few of them said it while actively being somewhere they didn't want to be.

So I built FireCoach.ai.

FireCoach loads every rep with your methodology before they ever get on a live call. They go in knowing the objections, knowing the right answers, knowing exactly how to deliver them.

Like having the test questions the night before.

🔸AI bots run the exact buyer conversations your reps will face, before they face them live

🔸Every session scored against your methodology (Gap Selling, SPIN, MEDDIC, or we build one with you if you're running off vibes and instinct, which honestly most teams are)

🔸Reps see exactly where they stumbled and what to say instead

🔸Managers get rep-level visibility into every rep's gaps without sitting on every call

🔸Works with any sales motion: inbound, outbound, enterprise, transactional

🔸Your reps don't walk into calls hoping they remember their training.

They walk in having already won the conversation.

Basically cheating. We're fine with that.

The part most tools skip: adoption. Every customer starts with a PowerSprint → we come in, customize FireCoach to your exact methodology, and we stay until your team is actually using it and getting results. Not a help doc. Not a Loom walkthrough. Us, there, making it stick.

Tool and team. You're not doing this alone.

I'll be in the comments all day with @thisiskp_ @yyogeshwar @vishal_maurya03. If you run a sales team, I want to hear what's actually broken. I'll be straight with you about whether FireCoach fixes it.

Try it free at firecoach.ai 🙏

this is cool, we’d use to train SDRs

Hey PH fam 👋

Thrilled to hunt FireCoach today. This problem statement resonated immediately.

As a sales leader and life long business owner, @midori_verity has seen the same revenue leak at every company she's touched. Not a people problem. Not a product problem. A readiness problem.

Most reps walk into their first real calls carrying half a playbook and good intentions. The coaching they needed happened in someone else's 1:1. The manager with the right instincts was already in 12 other meetings. By the time the gap shows up, it's showing up in the pipeline. So she teamed up with @yyogeshwar and built a solution for this: Firecoach.

FireCoach moves the coaching before the damage happens.

→ 🔸 Reps practice against AI that mirrors your actual buyer conversations

→ 🔸 Sessions scored against your methodology: Gap Selling, SPIN, MEDDIC, or built from scratch

→ 🔸 Every rep gets visibility into their own weak spots before a manager has to point them out

→ 🔸 Managers see rep-level patterns without adding a single meeting to their calendar

What I find most compelling is what happens after you sign up. Most tools onboard you and disappear. FireCoach runs a PowerSprint where the team comes in, customizes everything to your motion, and stays until adoption is real. That's a fundamentally different kind of commitment.

PS: I've enjoyed being an advisor to the team and truly believe they are a powerhouse combo!

The team is in the comments all day. If something in your sales pipeline is quietly bleeding, go tell em what it is. 🙏

Good product, 1 question though -
How do you keep the methodology scoring accurate—is it just keyword matching, or does it actually evaluate the quality of a rep's response?

The strongest part here is tying coaching back to real calls and outcomes, not just making roleplay feel realistic.

One thing I’d watch carefully: scorecards can accidentally reward reps for sounding like the methodology instead of actually moving the buyer forward. I’d want every coaching note to keep a source trail: the exact call snippet, the outcome it correlated with, and whether a manager agreed with the interpretation.

That would make the “clone the top performer” promise much more trustworthy, because the team can see what was copied, what was inferred, and what still needs human judgment.

Very cool @midori_verity! Upvoted :)

So how do you train AI on what's in people's heads? Do I need to record my sessions and AI trains on that? And how long will it take to replicate what my best sales person would do?

Congrats on your launch!! Do you tie the coaching insights back to CRM outcomes over time, or is it mainly performance inside the roleplay?

Nice idea, would be very beneficial for my team as well. What CRM can you integrate? I'm assuming SalesForce, how about SAP CRM?

Hi builders!

Keith here, CEO and co-founder of AppFactor. Really excited to ship this one.

The story
We built MCP Bridge for our own need, not an idea.
At AppFactor we've spent years building deterministic tools for an orchestration system of agents that deliver autonomous software maintenance. Infra and software discovery, scanning orchestration, build engines, deployment automation. As we layered our upcoming agentic platform (ForgeCatalyst) on top to harness these tools, we hit the wall every team building production agents eventually hits. Security, governance, cost/token usage, observability...
The AppFactor system requires meaningful validations, in environments where governance, security and controls are paramount when acting on customer code. With large complex API's with many tools, comes the next challenge - context constraints and efficiencies and multiple protocols to support.
The standard fix is to hand-build a dedicated MCP server for every API. This doesn't scale. We know we are not alone with our requirements and given the domain we operate in which is all about software maintenance, legacy transformation and the eternal battle of trying to bridge the past to the future. We recognised that not all systems have clean, well presented OpenAPI spec API's. There are many API protocols, and almost all API's were indeed built before agents and LLM's became an exciting real world proposition. That poses challenges in how these tools are invoked and consumed. So we built MCP Bridge to address all of these challenges.
You know the rest of the story.

What it does
Point it at any REST, GraphQL, SOAP, or gRPC API. It auto-generates fully-typed MCP tools with behavioral annotations and smart response processing. Self-hosted. Open source. Credentials never leave your environment.

What's shipped

• 4 API types, end-to-end

• 6 auth methods (Bearer, Basic, API key, OAuth2, Cognito SRP)

• Human-in-the-loop approval for destructive ops

• Code Mode: 3 meta-tools replace 100+ definitions, ~98% less context

• Analytics: latency, token cost, errors per tool

• Built in Rust (Dioxus + Axum), PostgreSQL, in a container

We're in the comments all day. What APIs would you connect first? And how should we improve?

Happy building!

Keith
→ MCP Bridge: https://mcp-bridge.ai/
→ Docs: https://docs.mcp-bridge.ai/

Thanks @fmerian for hunting us!

Most "connect any API" tools mean "connect any API that already has an OpenAPI spec and clean auth." What does MCP Bridge actually do when you're dealing with something messier, like a legacy REST endpoint with inconsistent error codes or an API that requires a multi-step auth handshake before you can do anything useful?

Congrats for the launch!

So happy to launch MCP Bridge! 🚀

Auto-generating typed MCP schemas from SOAP and gRPC alongside REST is the hard part others skip. We've spent cycles manually wrapping customer-side APIs with inconsistent auth patterns just to get an agent to call them reliably. How does schema inference hold up for APIs without an OpenAPI spec, or ones where the response shape varies by query?

the security breakdown in the comments is the most honest thing I've read on a PH launch. most tools in this space skip the 'what happens when your agent gets prompt injected' conversation entirely. the 3-level sensitivity guardrails and credential encryption is how this should be built

the auto-generated schema approach is smart. biggest pain point with MCP right now is writing tool definitions by hand for every API you want to connect. curious how it handles APIs with inconsistent response formats across endpoints?

We've run into this problem a few times while connecting different agent workflows. The integration itself is usually easy, but keeping everything reliable once multiple tools start talking to each other is where things get messy.

Are most teams using MCP Bridge as a central layer between agents, or more as a quick way to expose existing APIs to AI tools?

Finally, a product to help you create a MCP server, fast and reliable. It's about time! If you don't use @MCP Bridge by Appfactor, you're ngmi.

S/O for the launch, ?makers

This could save developers a lot of integration time.

How does auth work when the target API needs OAuth? that's the part that always kills these tools for me

Using MCP as the standard transport is a smart call. Agents get structured tool definitions without any custom adapter code. We've hit the N+1 connector problem repeatedly building AI agents that need to talk to CRMs and ticketing systems. How do you handle auth token lifecycle when agents discover and invoke new endpoints dynamically?

Congrats on your launch @keith_neilson @ehw_appfactor @knarik !

You provide any way to test the MCP tool definitions created as well?

How do you handle user permissions without leaking secrets to the agent?

"Any API" is the key claim does it handle APIs that require OAuth or custom auth flows, or mostly simple API key setups? That's usually where these tools hit a wall.

Hey PH! 👋

We’re the founders of Integuru, an AI-powered integration platform.

AI agents need to access external platforms, but most platforms don’t offer official, accessible, or comprehensive APIs. That makes integration the single biggest bottleneck.

To build integrations, most companies today use browser automation. But many of them have now outgrown their browser implementations. They’re running into issues with latency, reliability, and throughput.

Introducing Integuru: an AI platform that automatically reverse-engineers a website’s backend. It generates integrations that directly hit website servers. Today, our main users are vertical AI companies in healthcare, logistics, proptech, legal, and more.

On Integuru, you can generate your ideal integration end-to-end in minutes, including support for authentication. In real production environments for existing users, API calls complete in ~3 seconds with 99.9%+ success, including deployments above 1M requests/month/platform. Requests are executed live and achieve low latency without data-caching.

Integrations on Integuru get auto-healing, API docs, and a 24/7 on-call team for maintenance.

⚡️ Using Integuru is super simple:

1. Input the website and account you want to integrate with.

2. Describe your desired feature in natural language.

3. Integuru proposes an API schema and, upon your approval, generates the endpoint in ~10 minutes.

🙋‍♂️ Who is Integuru for?

Integuru is best for vertical AI companies that need to integrate with customer systems. We already support high-growth companies across these industries:

• Healthcare: connecting to EHRs, EMRs, RCMs, CTMSes, insurance portals, etc.

• Logistics: integrating with TMSes, WMSes, ERPs, carrier portals, etc.

• Proptech: automating property management systems, utilities, permitting software, etc.

• Legal: controlling practice/case management systems, government portals, etc.

• Enterprise AI transformation: accessing ERPs, CRMs, CX platforms, internal tools, etc.

… and more

🎯 Try Today!

Integuru is free to try! Go to app.integuru.com to get started! We’d love your feedback ❤️

congrats on the launch! I was wondering...brittleness in browser automation usually shows up much later on, when the platform ships a quiet backend change. So when that happens, what's the time-to-healed-endpoint your consumers actually see? Hours? Minutes?

And what if the required data is behind a login? Will the service be able to log in (if login/password are provided) and retrieve data from the personal account/dashboard?

@richardzhang Congrats on the launch! Solving the "no API" bottleneck is huge for vertical AI teams.

The promise of "reverse-engineering the backend" is powerful, but for healthcare/legal users (as you mentioned), security is paramount. How does Integuru securely handle authentication tokens and session management? And if the target website changes its backend structure, how does the "auto-healing" feature detect and fix the integration without breaking production data flows?

Congratulations on the launch! Replacing fragile browser automation with direct backend integrations is a bold move, and the reliability gains alone sound compelling.

the 24/7 on-call maintenance team detail is doing more work than it might seem. the hidden cost of browser automation isn't the initial build, it's the ongoing babysitting when things break at 2am. if that maintenance burden actually transfers to Integuru's team that's a meaningfully different value proposition than just being faster

That sounds interesting 🤔

Would love to know the architecture of it

Reverse-engineering internal APIs by watching network traffic is clever, but the thing I'd want to know is how fragile the output is when the platform rotates tokens or changes an endpoint path. Does Integuru detect that and regenerate, or does the caller just start getting 401s until someone reruns the capture?

Congratulations on your launch!

reverse engineering backend apis dynamically to bypass brittle puppeteer/playwright scripts is a wild approach. how are you guys handling complex, multi-step authentication flows like oauth rotations, recaptcha checkpoints, or enterprise sso under the hood when generating these endpoints?

browser automation is such a massive pain to maintain at scale with selector drift, so hitting the server directly is a huge upgrade. outstanding positioning on the launch team..🙌

Congrats on the launch, Richard! Framing integration as the bottleneck for agents really lands — browser automation has always felt like a fragile stopgap once you hit latency and reliability at scale. Turning that into clean APIs with ~3s calls and 99.9% success is exactly what agent builders have been missing. Following along — I'm @JayTheSong on X if you'd like to connect.

"No browsers" is a bold claim for platforms that actively fight scraping. How does it handle platforms that update their structure frequently or add bot detection? Curious how stable the generated APIs are over time.

Browser automation for integrations is one of those approaches that looks fine in a demo and becomes a maintenance burden in production. The moment a platform does a minor UI refresh, your agent is blind. We evaluated this class of tool when building integrations against APIs that didn't have official endpoints, and the failure mode that killed us wasn't the happy path, it was authentication session expiry mid-workflow with no clean recovery path. How does Integuru handle auth token refresh for platforms that use short-lived sessions or MFA flows? Does it abstract that entirely, or does that still land on the caller to manage?

This is awesome! I would love to use it. Though, quick question: How legal is it? Will the mega SaaS companies send us a cease and desist for using this? Can they detect that we are using it? Will they block our customers' accounts if our customers are using it?

Review code in @Linear, iterate with your team and agents, all in one place. Love this direction!

Upvoting and commenting so I can get that awesome swag! Totally kidding, Linear is great. Congrats on the launch, love the new swag—you did well.

Been waiting for something like this since the "close the issue, open GitHub, scroll to find the PR, lose context, come back" loop gets old fast. I do not use code editor anymore so a better interface that github is always welcomed! Very cool feature

I dont read diffs anymore tho, there are just too many prs. But I do ask 3 agents to read them on every PR and just give me the gist. Can it do smth like that?

Code review inside Linear makes a lot of sense now that agents can create PRs directly from issues. The hard part is keeping the review anchored to the original intent, not just the diff. Curious how you handle stacked or related PRs from the same Linear issue.

One thing we've seen with AI-generated code is that PR volume goes up faster than review capacity. Are teams using this primarily to speed up reviews, or to reduce the amount of context reviewers need to load before giving feedback?

I haven't used Linear, just out of curiousity, isn't it mostly a PM tool ? If so, why would we need to review PR here ? Shouldn't it be in the IDE / coding tool ? I didn't get the use case.

Rendering PR diffs inside the issue context solves a real context-switching problem. We've lost too much review state bouncing between GitHub threads and Linear tickets. Since reviews sync back to GitHub, curious how you're handling force-pushes mid-review. Does the diff view update automatically, or does it require a manual refresh?

with agents pushing PRs constantly now, having review right next to the issue is actually necessary not just convenient. nice timing on this

Collapsing the context switch between issue tracker and code review is the right call. You can finally review a change against the spec it came from without toggling tabs. We've felt this friction most when AI agents generate multiple interdependent PRs in a sprint. Does Linear Diffs support inline comments that sync bidirectionally with GitHub's review state?

I love it. Even that typing is so real!

Can it measure elements across different browser zoom levels?

Right thing shows up at the right time ! I'm making a pitch deck for my product and it's so hard to center / align stuffs and this help a lot. Well done G

Now that is something we all never asked for. But we all need it

Nailed it man 🫡

As a designer this seems useful to do design QA and provide specific feedback to devs on what to change.

I've typically done this manually through browser developer tools. I find the item, adjust the css until I get the desired result (when possible). Then I give that info to dev for updating.

This will save a bit of time, although it would nice to be able to edit the code from the Screen Ruler sidebar. Then I wouldn't have to use dev tools at all. But I'm not sure the browser allows that.

Great work!

Haven't launched on Product Hunt yet, but this caught my eye. Most rulers just measure. You added actual dev tools.

Genuinely curious – devs or designers as the main user?

We've found that users often ask for analytics but end up wanting answers rather than dashboards.

Are you seeing customers primarily embed charts, or are people starting to ask for more conversational/AI-driven ways to interact with their data?

Hi friends! Kris here, also from Basedash.

The before-and-after on this one still gets me. Adding analytics to your product used to mean a quarter of eng time and a permanent maintenance tax. Now it's an iframe and a JWT.

And your customers get the full product. The agent, dashboards they build themselves, insights that show up on their own, all scoped to their own data.

If you've ever had a customer ask for a dashboard inside your app and quietly winced, this is the one to try.

Drop your questions below, I'll be around all day.

BaseDash is so, so goated. @maxmusing built tech that made me fall in love with BI, again. The insights generated from just plugging in, like, PostHog & a couple others = insane.

I do, candidly, wish it were more affordable, or that they offered plans that were more accommodating of early stage companies, since I think those startups could benefit the most from the data surfaced by BaseDash, but I get it... it's definitely worth what they're charging, but the Basic plan paywalls so many of the useful features (MCP servers, full source support, etc.) that you're really looking at $1k/mo... maybe some sort of a program like Intercom & others offer, where a company can qualify for a special 1-2 year "growth startup" package, could be apropos?

Really excellent tech, nothing quite like it. It's worth running a trial if you have a SaaS product with even minimal traffic, and are using something like PostHog, MixPanel, Amplitude, Fathom, GA, databases, etc.

Keep up the great work, Max - I hope you consider opening up some non-technical roles on that /careers page someday! ;) (I'm gainfully/happily employed, mind you! but BD is one I'd consider joining...)

about the customization controls. you mention deciding which features customers see but how granular does that actually get. like can i show the dashboard view and hide the AI chat entirely for certain customer tiers, or is it more of an all-or-nothing toggle. that distinction matters a lot if you're trying to gate analytics features by plan without building a separate integration for each tier

How customizable are the embedded dashboards?

Hey PH 👋,

Hyper cofounder here. My cofounder Kanyes and I have been (best friends and) building together for last decade. We’ve been power users of nearly all the second brain / tool-for-thought tools as students back in 2017, then later tried fine-tuning GPT-2 in Kanyes’s garage back in 2020 to build an AI that knew everything about us — our taste, how we work, what we believe in, etc. and have that AI automate work that we found tedious (like doing problem sets, yuck). We were obviously 5 years too early.

After college, we joined Matic to build robots, and we again felt this pain but in a real company that needed to ship real products. We saw information constantly fall through the cracks and tribal knowledge lost when key people left the company. We watched people trying their best to prevent this by being disciplined with writing docs or posting slack threads, and eventually we became those people. Then in 2023 when the world shifted to AI agents, we realized that even thought we’ve hacked “good enough” solutions together for human workers, our AI workers still remain dumb! 

They’re as powerful as executives but have the context of a day-one intern. 

So we both quit our jobs and decided to go solve this properly. Today, I’m excited to share that you can download and start using Hyper directly from our website! We're deeply grateful to be able to deliver world class software to the dreamers and pirates like us. Hyper has already sped up our AI automations significantly (drafting emails that has all our company knowledge and writes like me!) and we're relentless about fighting hard for the world-bending teams that want to move extraordinarily fast.

Congrats on the launch!🚀

The support ticket triage use case is probably the most immediately valuable. That's a task with clear ground truth (was the triage correct?) and high volume, which means you can measure improvement fast. The harder version of that problem is when the institutional knowledge needed for correct triage lives in a decision someone made 18 months ago in a Slack thread that no one bookmarked. Curious whether Hyper indexes Slack history retroactively on setup, or whether it only captures context going forward from the point of connection. that distinction would significantly change the time-to-value for new customers.

This is super cool. Congrats on the launch guys!!

The company-brain framing is strong because it names the real bottleneck with agents: they can act, but they usually do not know what the company has already learned.

One thing I’d want to see very clearly in a product like this is how Hyper separates shared company truth from personal working style. “We always position the product this way” and “Shalin tends to write emails this way” are both useful, but they should probably age, conflict, and get overridden differently.

If the system can show source trails and confidence for those two layers separately, it would make the voice/email use cases much more trustworthy.

Pulled up the orbit visual and got why you went circular for scales. Symmetry that's invisible on a keyboard becomes obvious on a circle right? When you reharmonize a bar, does it stay diatonic by default?

The visual harmony approach looks very beginner friendly.

The visual map is a cool idea, but I'm confused on what it actually represents. Especially the second picture.

@adictum very interesting!
Do you have to be a pro musician, or you can figure it out even if you're an amateur?

Pretty cool i dig it. Your chord variations are using guitar tabs which i found a little weird for a piano app but as a player of both i can see the advantage there. In general its a cool product with some cool things to offer.

Im not an expert at marketing but maybe throwing the price around so much might hurt. The price point i think is fair its just that its mentioned everywhere on the site, but i do get that a lifetime is unheard of these days haha. Again i am no expert at all its just my opinion. As for the app itself the visuals are really great and a beautiful way of seeing all this theory. Good luck!

This looks surprisingly polished for a free tool. Wishing you a successful launch day and lots of happy Mac users!

Hi Vishva! four things fighting for the same tiny strip! when a timer's running and you copy something and a download finishes all at once, what wins the notch? hard priority order, most-recent, or does it stack and rotate? congrats on your launch, good luck!

how you're handling the performance side. notch apps that run constantly have a bad reputation for quietly draining battery on MacBooks, especially on M-series chips where people are very sensitive to anything that tanks their all-day runtime. have you done any benchmarking on idle CPU and battery impact because that's usually the reason people uninstall these after a week regardless of how useful the features are

That looks really good.

I think you should let users add their own tabs.

Another suggestion is to add any LLM tab

Does Notchy have any battery or performance impact when it’s running all day?

Congrats on launching Juan!
What's the auth model when the phone itself is the dev environment? A laptop being stolen means an attacker has to crack disk encryption, hit your password, etc. A phone gets unlocked dozens of times a day in public, and if MoDev is logged in and connected to GitHub + Supabase + Vercel + Gmail, "phone stolen" suddenly means "production access stolen." Is there a step-up auth before destructive actions (deploy, drop table, force-push), or session re-verification on suspicious activity, or is it relying on the phone OS lock as the security boundary?

How do you handle safety rails, like preventing accidental production deploys or risky database queries from chat?

Interesting concept, so it works directly with github ? where is the code executing ?

@jaygangwar thanks! On performance, the phone is just the chat + control layer. The heavy lifting (builds, deploys, queries) runs on your own Vercel/Supabase, not on-device, so it stays light even on mobile.

On Replit: Replit hosts your code and runs the IDE + compute in their cloud. MoDev is BYOK — it never hosts your code. It's the mobile-first AI cockpit on top of your own GitHub + Vercel + Supabase. Different category: we're the control layer, not the host.

@ferdisigona great framing — 'phone stolen = prod access stolen' is exactly the risk to design around.

Today: every destructive/write action needs an explicit in-app confirmation, and the worst ops (drop table, delete repo, charges) are hard-blocked entirely, so a logged-in, unlocked phone still can't deploy or wipe data silently. The current session boundary is OAuth + OS lock.

What's not there yet: a biometric step-up gate before destructive actions and session re-verification on suspicious activity. That's the next security layer on the roadmap. Not going to pretend it's shipped, it's where I'm headed.

@karimbenkeroum Hey Good morning! Great question, this was core to the design. Every tool is split into read vs. write. Reads (queries, listing repos/deployments) auto-execute. Anything that writes - deploys, DB mutations, sending mail - surfaces a confirmation card before it runs, so nothing fires silently from chat.

On top of that there's a hard-blocked tier that never executes even if you confirm: DROP TABLE, deleting repos, and live Stripe charges. The model literally can't run those.

And because it's BYOK, you're operating against your own GitHub/Vercel/Supabase, ETC. MoDev never hosts your code or holds the keys to do something irreversible behind your back. Human-in-the-loop on every destructive action, by design.

The ui looks really good not sure about the performance.

How will you differentiate from replit?

Wow, interesting, will check this out, I was looking for something like this. Does it support gemini/codex too ?

the budget tracking piece is interesting for teams but right now it sounds personal use only. is there any plan for shared visibility, like if a small team wanted to track aggregate spend across multiple machines without everyone having to check their own notch individually. or is keeping it single machine intentional because the moment you add sync you have to make privacy tradeoffs you've explicitly avoided

That is really good.

Does it track calude code/antigravity costs or usage

Love that it lives directly in the Mac notch.

The local-only processing is the right call for an API key-adjacent tool. Anything that proxies or phones home with usage data is a non-starter for a lot of developers. One thing I'm curious about: how does TrackNotch get read access to the token counts? Is it reading from the local provider SDKs' log files, intercepting at the network layer, or pulling from the official usage APIs on a poll interval? The answer affects both accuracy and whether it works for team accounts vs. individual API keys.

Hey, Tuan. I've got a question.

So, I was looking for a plan to travel from my city to Antarctica. There are many stays listed here with ratings. When I clicked on them, there were no reviews. So, I'm wondering where the ratings are scraped from. Is it from travel booking sites like Agoda, Booking.com and Aibnb?

Congrats on the launch! 🐰
As someone who's constantly hopping between cities for content and work, the "optimized itinerary" angle hits real. Most travel apps give you a list — this actually looks like it thinks for you.
Upvoted. Will test on my next trip. 🌊

pay as you go with no subscription is such a better model for creators who don't clip videos every single day. built the whole thing under air raid sirens in ukraine too. respect

No copy-paste and auto-delete? That's smart.
Genuinely curious – why Telegram instead of Discord or a web app? Either way, solid work.

Hi Product Hunt! 👋 I’m Serge, the solo-founder of Clipline.

I built this entire infrastructure over the last 5 months in Nikopol, Ukraine, often working under air raid sirens and power outages.

Why I built this:

As a creator, I was tired of manually cutting videos, re-typing subtitles, and waiting for renders. So I built a bot that does everything automatically – directly from the Share button. I wanted to create a tool that proves you don't need heavy desktop apps or expensive subscriptions to create high-quality content.

What is Clipline? 🎬

It’s a powerful Telegram bot that replaces tools like Opus Clip. You simply send a YouTube link or upload an MP4, and our AI (running on fast GPUs) does the rest in 90 seconds:

✨ AI-editing mode: finds the most viral hooks (no awkward cut-offs).

💬 Adds dynamic, engaging subtitles.

🌍 Auto-translates foreign videos into perfect English 🇺🇸.

🚩 Embeds your custom banner/watermark permanently.

🖐 PRO-editing mode: complete manual control - choose the pieces that will be hits.

✂️ Smart Remote Control: Unlike other AI tools that force random clips on you, our WebApp lets you manually select the exact 30 seconds you want to process.

💰 Pay As You Go: No month subscription traps. You only pay for the minutes you process using Telegram Stars.

🤝 The 20% Affiliate Program:

If you are a creator, marketer, or channel admin, you can monetize with us. We have an official 20% lifetime affiliate program via Telegram Stars. You earn a direct cut from every user you refer.

🎁 No promo code needed – every new user gets 5 free processing minutes right after /start. Just try it: send any YouTube link, and the AI will make shorts for you. The first render is on me.

I’m incredibly excited to finally share this with the PH community. I’ll be here all day answering your questions and reading your feedback! Let's make some viral clips! 🚀

Hey Serge! 'hold & refund' is a billing model i've not seen on many tools. In your case what trips the refund? clips you reject, ones you never download, or ones an AI gate flags as weak? and who judges 'what you got', you or the model? congrats on your launch, good luck!

We've noticed that once agents start calling multiple tools and sub-agents, debugging becomes harder than building the workflow itself.

Are you seeing people use PromptLayer mostly for observability after things break, or are teams actively using the traces to improve agent behavior during development?

Hey, Just came across PromptLayer and honestly it looks like something a lot of AI builders genuinely need right now.

One thing that caught my eye, the opening line "Observability for LLM apps" might be flying over alot of stressed users thinking "why did my AI just cost me $40 in one hour" or "why did it give that user a completely wrong answer"; not thinking in terms of observability.

I took a quick crack at a different angle for your hero section, happy to send it over if you want to take a look, no cost or anything, just thought it could be useful

Congrats on the launch!!

where did these tokens go' is the question every team running agents eventually asks and nobody has a good answer for. the waterfall view for multi-step workflows is what makes this useful over just checking your api dashboard

The waterfall visualization makes sense for sequential workflows, but agent loops introduce a wrinkle: the same step can execute dozens of times before terminating (or failing to). Does PromptLayer handle cycles in the execution graph, or does it assume DAG-shaped workflows? Specifically wondering whether it surfaces repeated-step patterns as a distinct signal rather than just summing token counts.

Wow this is like really really cool. Im sorry i have no use to try it as its just not in my wheelhouse of projects / products i work with but i was just looking at the products today and i was really impressed with this. I have seen a lot of like make this room look good with Ai but you guys went to a whole other level with this.

I definitely think this project should have been upvoted higher. I feel bad as i usually test the products i feel are cool and leave feedback but just dont know where to even start with this one. I did go through the site a bit though and some small recommendation's would be the landing page maybe should have some content or pics or video (this video you have here in ph is super compelling) but all these templates that if i click on take me to a login seems a little forceful. I just feel like its hard to learn exactly what you are about without being forced to sign up. I am no marketing expert what so ever, just wanted to point that out as its just my personal feelings. I thought it was just the product hunt slug page or something but i see your real homepage is the same. Anyway best of luck really amazing stuff!