PH热榜 | 2026-04-23

Hey PH 👋

YAN here, one of the makers behind Kollab. We built it so our team could stop bouncing between Slack, GitHub, Notion and half a dozen separate agent tools. One agent, sitting across every channel the team already lives in, with any MCP server wired behind it.

Here's how we use it ourselves. Kollab's hooked into our Slack and Telegram bots, with Notion MCP and GitHub MCP behind them. Inside our work group, anyone (devs or not) can ping the bot to look at code, review a feature, or file an issue. In the community group, users @Kollab to report bugs or ask how something works, and every message routes through Notion MCP straight onto our backend board. Feedback used to get lost in DMs; now it doesn't.

The piece we underestimated most is scheduled tasks. We thought we were shipping a digest job, but a scheduled task on Kollab is really a timed agent. The same cron can call any MCP tool, pull from the knowledge base, run as a specific agent role, and post back to any channel. Ours right now: one drafts a weekly changelog from GitHub issues, one cross-checks our status page against Sentry, one pings the on-call before standup. Same thing under the hood, totally different jobs on top.

When we need more than a quick answer, there's AgentCore. Long-running agent with its own filesystem and a browser built in. We've been using it to stand up small demo sites and internal tools instead of writing throwaway scripts. And since skills are just regular GitHub repos, anything the team keeps repeating turns into a skill the whole org can install by name. We're still early on this part, and it's probably where we'll end up finding the weirdest uses.

Question for PH: if you had one agent sitting across your team's channels with full MCP reach, what's the first scheduled task or skill you'd write? No idea what people will come up with. So far the answers have been all over the map, and two of them are already in our next release.

Hi Product Hunt! 👋 I'm Gavin, the CEO and founder of Kollab.

While building my previous SaaS product (Buildin), I realized a fundamental issue: even with deep AI integration, most tools operate on a "SaaS + AI" logic where AI is merely a helpful sidekick. However, with the rapid rise of Claude Code, MCP, and similar breakthroughs, we are officially entering the Agent era.

Yet, the barrier to entry for using Agents at work is still way too high. Terminals, npm installs, MCP configurations, system prompts, memory management... these technical hurdles keep 90% of everyday users out. Even for the tech-savvy who do know how to set them up, their Agent environments remain siloed on local machines, making it incredibly hard to share workflows or best practices across a team.

That’s exactly why we built Kollab. We designed Kollab to be the central hub for team-agent collaboration. We focused on three core pillars to make this happen:

1. Zero-Barrier Configuration: We made the complexity of MCPs and coding environments completely invisible. Through our Connectors, you can integrate tools like Notion, GitHub, Figma, Linear, and Slack with just a few clicks, allowing your Agents to seamlessly access and act on your actual business data.

2. The Compounding Power of Team Knowledge: This is what makes Kollab truly special. When any team member creates a new Skill or sets up a workflow, it’s immediately added to your team's shared Skill Marketplace. One person's "aha" moment instantly scales into an organizational capability. No more reinventing the wheel.

3. Work Where Collaboration Already Happens: You shouldn't have to change your habits to use AI. With Kollab, you can deploy your Agents as Bots directly into Slack or Telegram. Just tag them in your chat, and they’ll take instructions and execute long-running automated tasks right alongside your human teammates.

Internally, our product, engineering, and ops teams are already sharing over 20 active skills for our daily workflows. We firmly believe that Agents shouldn't just be about boosting individual productivity—they should serve as the central nervous system for team collaboration.

We’d love for you to try Kollab and would be incredibly grateful for your honest feedback!

👉 https://kollab.im/product

Hey 👋 I'm jiayi, one of the makers behind Kollab.

Kollab is an AI-native workspace. Unlike doc tools with AI added on top, Kollab puts Agents front and center. You give them tasks, they execute, and everything stays in a shared workspace your team can actually use.

Here's a real example. Our team runs a blog. It used to be all manual: track trends, find topics, write drafts, make images, review. Same grind every week.

Now in Kollab:

1. A scheduled task searches target keywords every morning and drops new topic ideas into the workspace

2. Another task picks up new topics automatically, writes drafts and generates images

3. A review task runs a saved Skill to check tone, structure, and SEO

4. When it's done, the Bot sends a message in our channel so the team knows it's ready for final review
   

Three scheduled tasks running in the background. Skills defined once, reused every time. We just do the last step: review and publish. What used to take a team days now takes one person a few minutes.

No code. No stitching five tools together. Set up a Skill, set a schedule, let Agents do the work.

Teamwork, done with Kollab.

We've been using Kollab internally for a few weeks now.

The biggest win for us is Skills — once someone builds a workflow, the whole team can reuse it instantly. No more explaining the same process over and over. Really changes how we share knowledge across the team.

This feels very practical. Most teams don’t lack tools — they lack something that ties everything together. An agent that sits across channels and actually executes workflows (not just answers) could remove a lot of operational overhead.

Hey 👋 I'm Lynn, one of the makers behind Kollab.

Kollab is the AI workspace that actually gets how teams work. 🎯I've been looking for a platform that doesn't just bolt AI onto project management, but truly unifies agents, knowledge, and team collaboration in one place. Kollab nails it.

What stands out:

• One CLI to rule them all — spaces, projects, tasks, skills, bots, timers, MCP servers, memory. Everything is accessible through a single, cleankollab command. No more jumping between ten different tools.

• Knowledge-base powered — ask questions across your projects and get real answers grounded in your docs, not generic LLM hallucinations.

• Agent-first by design — timers, bots, and skills aren't afterthoughts. They're first-class citizens you can configure, automate, and deploy.

• Model flexibility — choose between Lite, Pro, and Max tiers depending on the task, so you're not overpaying for simple queries.

For teams building with AI, Kollab feels like the operating system we should have had all along. Clean architecture, real automation, and collaborative by default.

Upvoted and excited to see where this goes! 🚀---

Interesting positioning. Feels less like “another agent tool” and more like an orchestration layer across where work already happens. If teams can actually rely on it for day-to-day ops, this could become pretty sticky.

Didn't expect this one to land for me, but it did.

The core bet is that the agent should live in Slack or Telegram instead of some separate dashboard you have to open. That's just correct. Most teams aren't lacking AI tools — they're lacking time to go find them when they actually need them.

The Skills system is what shifts it from "team chatbot" to something real. One person builds the workflow, everyone reuses it. Luo from HeyForm said it better than I can: no more explaining the same process over and over.

One thing I'd love to know: what's running under the hood, model-wise? And is there a path to bringing your own API key? For teams that already have Claude or GPT-4 access through work, that could be a dealbreaker — or a non-issue, depending on how it's built.

Also curious about MCP tool limits. ChatGPT caps at 30 tools per connector — what's the ceiling here? With complex workflows pulling from GitHub, Sentry, Slack, and a few others at once, that number matters more than it looks. Upvoted!

I've tried a bunch of AI productivity tools, and most of them feel like single-player experiences. Kollab is the first one that actually makes sense for a team.

The bots triggering from Slack and syncing back to the workspace is a nice loop, but how about when someone edits the output inside Kollab, does that change reflect back in the Slack thread?

Love seeing this live — adding agents into team workflows is a smart, natural evolution. Congratulations on the rollout!

Super excited to see this live! Bringing agents directly into team workflows feels like the right layer to build on. Congrats on the launch!

it's great to have AI team! Congrats on this launch!

Congrats on the PH launch, Kollab! 🎉 @yan_labs_
Bringing AI agents right into Slack (without switching apps) + persistent Memory + reusable Skills = finally a workspace that doesn't fight the way teams actually work. 👏

Love the "scheduled task as a timed agent" idea – that's way more powerful than boring cron jobs. And AgentCore with its own filesystem/browser? Seriously cool.

One practical suggestion from a collaboration perspective: as teams scale trust in agents, consider adding role‑based permission templates for Skills – e.g., a Skill can read Notion but not write, or only usable in certain channels. That would lower the "what if the agent messes up" fear and unlock wider adoption. 🙌

Question for you: do you support custom agent personas right now? Like a "code reviewer" vs "customer support" bot with different tone/knowledge bases.

Congrats again – can't wait to see what wild scheduled tasks your community builds! 🚀

the Skills concept is interesting — is a Skill basically a reusable prompt+tool bundle, or does it carry its own memory/state across runs? trying to understand where it sits between a workflow and a full agent

I'm not a developer, and that's exactly why Kollab works for me. No terminal, no config files.

I just connect my tools, set up what I need, and the agent handles it. Finally an AI tool that doesn't assume everyone can code.

How do you handle agent coordination across workflows? Building an AI scheduling assistant for TV and curious about your approach to chaining agent tasks.

Hey Product Hunters! 👋

We’re really happy to be back here 2 years after our first launch, this time with FocuSee 2.0.

Over the past two years, we’ve kept working on one idea: making “record your screen and get a polished video automatically” feel more complete, more natural, and more aligned with how people actually create today.

Along the way, we talked with many users and learned the same thing over and over: people don’t just want to record their screens. They want to quickly turn rough recordings into polished videos for product demos, tutorials, walkthroughs, online courses, marketing videos, and more — without spending hours editing afterward.

That’s really what shaped FocuSee 2.0.

We kept refining the product around real workflows and habits, and this release is the result of that work.

What’s new in FocuSee 2.0?

✔️ Record mobile demos and tutorials as easily as desktop workflows, with iOS and Android screen recording now built in.

✔️ Go beyond flat zoom with 3D Motion, adding more cinematic depth when you want stronger visual storytelling.

✔️ Keep your delivery natural and confident with in-built teleprompter, so you spend less time stopping and starting over.

✔️ Make every step easier to follow with annotations, magnifier and highlight effects, auto subtitles, and visualized keyboard shortcuts.

✔️ Keep sensitive information out of view more easily with blur effects.

✔️ Make rough recordings look and sound more refined with AI background removal or replacement, voice enhancement, noise reduction, and filler word & silence removal.

✔️ Choose the presentation style that best fits your story with AI avatars and dynamic camera layouts.

All of these improvements are built around the same idea: helping you create more polished, professional videos without making your workflow heavier.

We’re genuinely grateful to everyone who shared feedback, feature requests, and real use cases with us over the past two years. FocuSee 2.0 is a much better product because of those conversations.

🎁 As a thank-you to the Product Hunt community, we’re also offering an extra 30% off.

Use code PH30OFF at checkout (valid for the next 3 days).

We’d love to hear what you think and answer any questions. Thanks so much for checking out FocuSee 2.0!

The gap between 'I just recorded this' and 'this looks professionally edited' has always been the biggest friction point for indie makers. Focusee 2.0 seems to nail that last mile especially the filler word removal, that one's a dealbreaker for me with AI voice tools since I'm not a native english speaker lol. The 3D motion is a nice touch too, adds that cinematic feel without having to touch a timeline

Congrats on the launch! 👏

The iOS screen recording built in is the feature I've been waiting for, as an indie iOS maker I've been jumping between QuickTime, my iPhone, and a separate editor just to make a decent app demo.

Two years of user feedback clearly shaped this into something much more complete.

Does the mobile recording work wirelessly or does it need a USB connection to the Mac?

Looks like a great all-in-one for SAAS demos. No need to use 10 apps for a single video anymore?

Hey PH family! 🙌 Co-maker here.

Morgan said it best, but I just want to jump in and add a massive THANK YOU to this community. FocuSee 2.0 was quite literally shaped by the feedback, feature requests, and support we received from our first launch here.

We're pushing the boundaries of what "zero-editing" means, and the new AI enhancements and 3D camera motions are designed to make your videos look like you hired a professional editor.

We built FocuSee 2.0 to do the heavy lifting for you so you can focus on your message, not the timeline.

We're all ears today! What kind of videos are you all planning to create with FocuSee 2.0? Let us know, and drop any questions you have below! 👇

Is it possible to add a device frame to the video

will give it a test today. Hoping its good - this is a real pain for someone with no video editing skills 😂

Does it give you the option to edit raw or lightly edited clips to your own timing in a main stream editor?

Hey Product Hunt! We're back with our 5th launch ever!

Today, we're launching Magic Patterns Agent 2.0, the culmination of 3 years of observing real customer workflows and AI design use cases.

Every day, a new AI design tool comes out, but Magic Patterns is battle-tested. Our first launch ever was 3 years ago on this website, and today 1,500+ product teams at companies like Granola, Vanta, and Freedom Mortgage use Magic Patterns to go from idea to production.

With Magic Patterns, designers, PMs, and founders use their company's real design system to create interactive mockups and hand off to engineering.

Today, on Magic Patterns, we're announcing:

• Skills: drop in any SKILL file to give our AI agent specialized instructions on demand

• Connectors: pull in context from your favorite tools

• MCP 2.0: export production-ready code to your coding agents with our MCP.

We've also made several improvements to our Agent. In our eval harness, we're seeing it use 15% fewer credits, 10% faster time-to-first-token, and an 8% performance improvement (measured by error rate). Agent 2.0 gives you faster, more reliable outputs at lower cost.

We can't wait to see what you'll build with Magic Patterns. Happy prompting!

P.S. It feels so good to be back on Product Hunt!

Daily user of MagicPatterns here. It has (NEARLY) completely replaced my need for Figma. I'm ALSO a heavy user of Claude Code - but there's nothing like going from one single prompt to a visual working prototype. And no tool does it as well as MagicPatterns - for the product designers workflow(IMO). I'm also a Lovable user - but for different use cases. When I'm in Design Mode -> I choose MagicPatterns. When I'm in - hack on this backend/scripting problem -> Claude Code. When I want to whip up a full stack app fast that I think I'll take to market -> probably Lovable. But again - besides MAYBE v0 - no one has been at this AS long - and with AS much of a customer obsession as Alex and Teddy at MagicPatterns. Big big fan.

Absolutely recommend magic patterns from the bottom of my wallet ❤️

The ROI of having a tool like this in the teams' pocket is immense. Your imagination is the limit. It's not only prototypes, but videos,

https://www.linkedin.com/posts/jorgeakairos_product-context-gets-fragmented-fast-calls-ugcPost-7453103110559989760-RLi1

local tools,

https://www.magicpatterns.com/c/5yywrafwujevlik1ljzqp7/preview

and even games...
https://project-anarchy-cookbook.magicpatterns.app/

This team ships 🚀

And they're also great people. Los máximos!

Product Hunt is where it all started for us, so excited to be back with this launch!

One thing I'm especially proud of: alongside Agent 2.0, we rebuilt the editor UI from scratch.

For those of you who've used Magic Patterns before, you'll feel the difference immediately:

• The chat thread is way easier to scan, surfacing the information that matters to you

• Visual Editing got a huge refresh, making one of our most popular features better and easier to use

• Model Picker is now available, you can now pick the right model for your prompt, all the way from Gemini 3 to Opus 4.7

Three years of watching how people actually use this tool went into this redesign.

We can't wait to hear your feedback!

Congrats on the launch I've been using it for my sales decks recently!

the "match your existing design system" angle is a huge deal — most design agents spit out generic tailwind and call it a day. how does the agent actually pick up on an existing system? scanning components, reading tokens, or something else?

How do you get the exact icons and colour scheme right? And how much data quantity is required from company's design.

I originally found Magic Patterns on Product Hunt mid-2024 and been building with it ever since.

I'm so excited to start using the different connectors with faster generations in Agent 2.0!

Let’s go team!

Really cool launch! Btw whats your moat over claude design or google's stitch? Is it affordability, speed or the actual design? Or correct me if am wrong and this is not even competing stitch or claude design

the handoff-to-engineering angle is the interesting part — does it actually respect an existing component library, or does it approximate the look and leave you to re-wire things? curious how that part works in practice

YES!

Cannot wait to give the agent a test drive with my OpenClaw setup. Congrats on getting this out there.

Hey what did you change in MCP 2.0 - how did the improvements help?

Hey PH 👋 I'm Shengkun, co-founder of Monid.

I've been building my own agents to automate my work, but it quickly became a nightmare. Connecting to 10-20 APIs, juggling separate subscriptions, payments scattered everywhere. I couldn't even track what I was actually spending.

So we built Monid, a wallet for agents to buy any paid tool they need.
One single balance. 215+ endpoints. No subscriptions.
Agents need flexibility and should never be blocked by paywalls.

With Monid, your agent can tap into powerful premium tools to:

• Access every major social platform including TikTok, Facebook, Instagram, X, LinkedIn, Reddit, and more

• Search for people across platforms with enriched details

• Pull blockchain data

• Pull ecommerce data

Our beta users are already using it for competitor tracking, social media content creation, VC sourcing, ecommerce product selection, and more.

Monid works with any agent, including Claude Code, OpenClaw, and Hermes Agent. Just send this to your agent to get started:

set up https://monid.ai/SKILL.md

Everyone here gets $3 in free credits.

We'll be hanging around all day to answer questions, please try it out and let me know how it works for you!

Hi, I’m Feiyou, co-founder of Monid 🚀

The promise of agents is autonomy.
But today, they’re still heavily constrained.

Without a human, an agent can only go as far as free tools and pre-built integrations. It’s limited to what’s already wired in.

With a human, the human becomes the bottleneck.

And often, humans don’t even know what tools or data sources exist.

So either way, agents hit a ceiling.

We believe agents need the ability to:

• discover tools and data at runtime

• decide what’s worth using

• and pay for premium resources when it makes sense

That’s why we built Monid, a wallet where agents can find and use the right tools on demand, including paid APIs, data, and services.

What’s something your agent should be able to do today, but can’t

Would love your thoughts, ideas, or use cases.

pay per call is how all API access should work for agents. subscriptions assume humans. agents are different.

the SKILL.md onboarding is the smoothest agent setup I've ever done. other tools should copy this.

how granular is the social media data? like can my agent pull comments or just posts?

Most 'agent platforms' are just LLM wrappers. Monid is actual infrastructure. Respect.

the API subscription model is broken for agent use cases. pay per call is the way.

x402 under the hood is a cool call — how do you handle the agent accidentally burning through a balance on a loop? is there per-tool spend limits or just a hard wallet cap?

Finally! An @x402 -based product!

the "no subscriptions, no API keys" framing is interesting — managing a dozen tool subscriptions for one agent gets expensive fast. how does pricing work under the hood, is it usage-based per tool call or some kind of credit system?

Super interesting! What do the safeguards look like to make sure my agent won't bankrupt me?

Very useful for the future boom of A2A interactions! Just curious whether it would be possible to set a ceiling/warning for agent expenditure. Would appreciate if the agents wouldn't drain my coffer too much before I know lol

My ecommerce agent now picks winning products autonomously because it can actually pull real marketplace data. Wild.

That's actually useful. Would users have the option to choose which tools they gonna buy for certain scraping tasks. like there might be a couple tools with price variation.

A single-pass code review, automated or manual, can only catch what one pass catches. Ultrareview takes a different approach.

It is a /ultrareview command for Claude Code that spins up a fleet of reviewer agents in a remote cloud sandbox, runs them in parallel across your diff, and independently verifies each finding before reporting it. The result is a short list of confirmed bugs rather than a long list of suggestions to triage.

The workflow is non-blocking by design. You confirm the review scope in a dialog, the agents run in the background, and findings come back as a notification in your CLI session when complete. Typically 10 to 20 minutes. You can close the terminal and it keeps running.

Key features:

• Multi-agent parallel exploration of the diff

• Independent reproduction step cuts false positives before findings land

• Remote sandbox keeps your local session free during the review

• PR mode pulls directly from GitHub, no local bundling required

• Each finding includes file location and fix context

Who it's for: Claude Code users on Pro or Max plans, specifically before merging substantial changes where a missed bug is expensive. Auth flows, schema migrations, critical refactors.

Research preview, available in Claude Code v2.1.86 and later. Pro and Max users each get 3 free runs to try it.

P.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends

This could significantly speed up PR cycle it false positives stay low.

been doing this manually — agent reviews, second agent verifies its findings, sometimes a third when something looks uncertain. you get closer to correct, but you're rebuilding the same scaffolding on every pr. does it actually catch what the third pass was catching?

parallel reviewer agents in a sandbox that independently verify before flagging is a nice take on hallucination control. one question — how do you dedupe when two agents surface the same bug with slightly different framings? always a pain with multi-agent review.

How does /utrareview handle reviewing integration issues? does it require any documentation on CI/CD setup and external dependencies?

Worth every penny

The 'all my videos look the same' feedback is exactly the right problem to solve, generic AI video is becoming wallpaper that nobody notices.

As an indie maker about to start creating content for my app, the Clay style caught my eye specifically for App Store preview videos.

Does Reloop handle portrait/vertical format well for mobile-first content, or is it optimized for landscape?

Conversation format can help in brainstorming ideas for campaigns but how much will be the input from tool's side. That would be the key to improving performance

tried to try it out.. I don't want to sign up and give credit card. just to try the software.

I don’t like a ChatBot who judges genuinely your answers in onboarding. Prefer to pass.

Did you license these styles from Pixar?

Hi Product Hunt 👋

AI tools have become powerful. But most of them still feel disconnected. You ask. They answer. The context resets. Nothing carries forward.

We built ASI:One to move beyond that.

ASI:One is a personal AI system that remembers, adapts, collaborates, and takes action. It is designed to stay with you over time, not just respond to isolated prompts.

No complicated installations on PC or cloud.

Here’s what you can do with it:

🧠 Build a personal AI that evolves with you

Shape its personality, set preferences, and let it remember what matters.

👥 Create Group Chats with AI built in

Invite others by email and let the AI help coordinate discussions in real time with friends and colleagues.

🗂️ Launch structured Collabs

Set a clear objective and let the AI break it into steps, track progress, and keep context intact.

🤖 Agents on demand! Type @agent inside a conversation and bring in domain-specific capabilities from Agentverse instantly.

📅 Connect Google Calendar and Gmail

Schedule events and handle follow-ups directly from your workspace without jumping between tools.

Under the hood, ASI:One routes tasks across multiple models and agents depending on what you are trying to accomplish. Research, planning, scheduling, coordination, and execution happen in one place.

We designed ASI:One for people who want an AI that works with them long term, not just for a single conversation.

We are here all day to answer questions and hear your thoughts 🙌

I’ve tested a few products with the same idea of a personal AI agent, and I always put them through a few real-world checks, especially email and daily task handling. A lot of them tend to fall short, or they simply don’t follow through consistently.

So far, ASI:One is passing those tests, and that’s pretty impressive. It feels like a strong product.

One question I had: when do you plan to start charging? I’ve been using it quite a bit and testing it a lot, but I haven’t run into any limits yet.

Overall, this looks like a great tool. Congrats to the team on the launch, and wishing ASI:One a great launch! 🚀

A personal AI with persistent memory that actually plans and acts is exactly where I see the most underused signal — markets. Prediction markets like Polymarket carry leading indicators that retail and institutional players both ignore because the data is noisy. I built PolyMind (https://polyminds.netlify.app/) to surface AI-driven alerts on the largest Polymarket trades in real time. Curious whether ASI:One's planning layer can plug into external probability streams like that.

How does it handle recurring tasks or scheduling around a user's calendar? Building something that schedules TV around when people are actually free and curious how you approach the planning layer.

Group chats with my friends their AIs: how does it handle privacy and sensitive information?

Congrats @rishankjhavar
BTW, how curated is the Agentverse network right now? Are agents verified in some way or is it more open marketplace?

I have some questions just think that I'm kind of thinking about I apologize if they come across blunt it's not my intention, but I'm seeing a lot of the same things being deployed and I'm trying to figure out what makes them different.

Is this just another implementation of Open Claw? What prevents a user from configuring this on their own? What makes yours unique?

How do you address the publicly documented issues and concerns?

I've seen so many iterations of this lately. What makes you guys really stand out in your opinion?

the "collaborates with others' AIs" part is what I'm curious about — is that agent-to-agent handoff happening in plain language, or is there a protocol underneath? feels like the messy part everyone handwaves past

Hey Product Hunt — Kai here, founder of Blink.new.

The reason we built Blink AI CFO is simple: I wanted an AI that trades stocks, builds the financial models, runs my projections, keeps the P&L current, and ships the decks. That's the whole job of a CFO boiled down to the parts I never wanted to spend my own time on, and no single off-the-shelf tool actually handled any of it end-to-end.

So we built the agent that does. Blink AI CFO autonomously executes stock and options trades through Alpaca or Robinhood on your pre-set rules, builds Wall Street-grade financial models, updates your P&L from live Stripe, QuickBooks, and Brex data, runs your projections, and ships investor decks on demand — all from a single message in Slack, Telegram, or Discord.

Every output is a real artifact, not a chat summary. Three things I'm proudest of:

→ Autonomous trading that actually executes. One message — "analyze the top 10 stocks by market cap and execute the trades" — runs the full research and places the orders through Alpaca, within your rules and with a human confirmation before anything live.
→ Wall Street-grade modeling on demand. Ask it to build a model for a public comp like MSFT, or for your own company's next quarter, and you get a formatted Excel or Google Sheets model with live formulas, scenarios, and charts — not a rough spreadsheet sketch.
→ Live P&L and projections. It pulls Stripe, QuickBooks, Brex, and your bank every day, keeps your P&L current, and rolls projections forward so your board numbers are always ready.

Under the hood you get Claude Opus 4.7, GPT-5.4, Gemini 3.1 Pro, and 180+ other models included with no API keys, plus 200+ connectors spanning Alpaca, Robinhood, Stripe, QuickBooks, Brex, Google Sheets, Slides, Docs, Gmail, and more. Live in under 60 seconds, from $22 a month.

Genuinely curious: what would you put an AI CFO to work on first — a trade, a financial model, a P&L update, or an investor deck?

I'll be responding to every comment today. Thanks for taking a look.

— Kai

Congratulations!

Having worked closely with this, the thing that genuinely surprised me was the data accuracy. For tasks this sensitive, that's always been the hardest problem to solve. Most AI tools will confidently hallucinate a number and keep moving. Gerald doesn't do that. When it's unsure about a data point it has sourced, it flags it rather than guessing, which is exactly the behaviour you want when you're dealing with financials that end up in front of investors or a board. It is also extremely crucial for automated trading through Alpaca/Robinhood.

The Excel and PPT output quality is also something you have to see to believe. The spreadsheets aren't just populated templates. Cells are actually linked, formulas reference each other correctly, and the model holds together structurally the way a real analyst would build it. The PPT design is equally impressive, the formatting, layout, and visual hierarchy are genuinely presentation-ready out of the box.

These are the two areas where most AI finance tools fall flat and they happen to be where Gerald is strongest. Really proud of what the team has built here.

Congrats on the launch! 🚀
The self-correcting bugs feature alone sold me. Every other tool gets you 80% there and then ghosts you.

Hey Kai! Looks awesome. Agents handling money could be a huge time saver. Wish you all the best here

I'm impressed by Wellow's approach to GEO — offering clear and actionable steps to take over an extended period (e.g. 90 days) of time to win greater share of citations on platforms like ChatGPT and Perplexity.

The pricing is also very reasonable.

What I particularly appreciate is their focus on leveling up your existing content rather than writing new stuff.

And yes, if you have gaps, they have a content engine that will help you fill them in, on brand, and on time!

Hey Product Hunt 👋

I’m Masab, founder of Wellows. Before I say anything else: we’ve launched on PH before — our previous product KIVA hit #3 in January 2025. So I know this community doesn’t mess around, and I don’t plan to waste your time.

The problem in one sentence:

Most AI visibility tools tell you if your brand got mentioned. None of them tell you why — or what to do about it.

AI models don’t pull answers from thin air. They pull from sources. If your brand isn’t in the sources being cited — Forbes, Reddit, TechRadar, niche blogs — your “visibility score” is borrowed, not earned. And your competitors might own more of that real estate than you think.

What Wellows does differently:

We go a layer deeper. When an LLM cites a third-party source, we scrape that citation and check whether your brand (and your competitors’) appears inside it. That’s what powers your AI Visibility Score — not raw mentions, but actual citation presence.

Then we close the loop:

• Optimize — We tell you which existing pages to point at which prompts before suggesting you create anything new

• Create — When new content is needed, our AI writing agent drafts it in a citation-ready structure

• Outreach — We surface the exact URLs where LLMs are already pulling from in your space, plus contact emails to actually get placed there

Customers like Cloudways, Secure, Vettio, and PureVPN are already using it. 6,000+ users have shaped every feature you see today.

A candid note:

Wellows is a work in progress — and I mean that genuinely, not as a hedge. Everything we’ve built has come from users telling us what’s broken. If something feels off, I want to know. I’ll be here all day.

PH-exclusive offer: 20% off all plans for your first 3 months — use "ProductHunt" when checking out.

And a huge thank you to @chrismessina for believing in this early and bringing us to this community. Means more than you know.

— Masab

Congrats on the launch. I have used Wellows and as Chris said, the unique thing about it is that it doesn't stop at just showing visibility. It exactly tells you step by step how to win on different AI platforms, specifically around content. I have known the team for a long time; it all comes from their own experience of running multiple sites that generated millions of monthly organic website visitors.

One advice I would give is to build connectors with tools like Buzzstream because the next step after creating content is to outreach and publish at the right places.

Been using Wellows for a while now, and the part that really changed how I think about AI visibility is the citation/source layer. Most tools stop at “you were mentioned X times in ChatGPT.” Wellows goes one step deepe. It shows which third-party pages the model is actually pulling from, and whether your brand is present there or not. That shift is important because it makes things actionable. Instead of asking “how do I get mentioned more,” it becomes “which of these 10–15 articles do I need to be part of?” That’s something you can actually work on, reach out, get included, improve positioning.

Also, the “optimize existing content first” flow helped a lot.
Congrats on the launch, Masab and team well deserved 👏

this is the angle most e-com brands are completely blind to right now. they optimize for Google but have zero visibility on how ChatGPT or Perplexity describe their products to buyers.

curious about the agency side, any plans for a multi-brand dashboard? agencies managing 10+ clients would need that before adopting.

The 'leveling up existing content rather than writing new stuff' angle is what got me, just launched my first iOS app and I have a landing page already, just don't know if it's AI-citation-ready.

Does Wellows work for mobile app pages or is it mainly focused on SaaS?

been tracking llm visibility for our own product manually —  the part that's hardest to do by hand isn't the initial snapshot.  it's knowing whether what you changed last month actually moved anything. the trend is more useful than the score.

Hey Product Hunt 👋
I'm Janis, CTO at Typewise, and it honestly feels a bit surreal to finally put this out into the world.

One of the things I'm most proud of is the UX. We didn't want another tool that requires a week of onboarding and a 40-page manual. Setting up Typewise should feel closer to onboarding a new team member than configuring enterprise software, you tell it what it should and shouldn't do, point it at the systems it needs, and it gets to work.

The hard part was making that feel effortless without cutting corners on safety. We spent a huge amount of time on the layers underneath: what the AI is allowed to know, what it's allowed to do, where a human needs to stay in the loop. That interplay between AI and human was the real design challenge for us, not the model itself.

I genuinely believe customer service is one of the clearest cases where AI can take the grind out of the job without replacing the people who care about it. And seeing it actually land with real teams is the part I still can't quite get used to.

After such a long build, I'm really curious to hear what you think of it. Honest first impressions, what stands out, what feels off, all of it very welcome.

Hey PH! I'm David, co-founder of Typewise. Here's a hot take: most "AI customer service" tools are just glorified chatbots with better marketing. They still need you to build flows, write rules, and babysit them.

We built Typewise to be fundamentally different. It's an AI agent system that builds your AI customer service for you. Describe your goals in natural language, and our platform creates specialist agents, connects to your systems, and starts resolving tickets autonomously.

But it also knows its limits. The AI supervisor detects when a human should step in and hands off seamlessly with full context. Your team picks up in a clean, beautiful UI that makes customer service actually enjoyable.

No flowcharts. No code. No manual tuning. The AI manages itself.

We're already live with enterprises and YC startups across Europe & US, and the early results speak for themselves. Excited to share this with the PH community. Would love to hear what you think!

I’m really proud to have been working on this as an AI engineer. I’ve been dogfooding it at my side company and it's become the thing I'd miss most if it disappeared. It’s saved me so much time responding to customer emails. Tickets get solved automatically with relevant context and tools, but also there’s a human in the loop element for tickets the AI can’t fully solve.

What about initiating refunds, e.g. we need to make them via Postman or in the Google Play Order management. Can it handle? Operate on cross-platforms?

One thing I've learned working on agents in production: what kills you isn't bugs, it's drift. Three weeks in, CSAT dips a few points and nobody on the team can actually tell you what changed.

That's why on the AI side we pushed hard to have simulation and evaluation as first-class primitives, not bolted-on features. It's the piece of this release I'm most confident about!

Going MCP-native from day one saved us from the usual “we’ll add that integration next quarter” loop. And with 6000+ connectors most customer requests are just a quick config change instead of something we have to slot into the roadmap.
From the engineering side, that’s what made this launch feel steady instead of a sprint we’d regret later.

Could be very useful! Honestly, I am so frustrated with some chatbots at many companies that provide no resolution at all. One question though if users don’t define flows then where does control actually live like in prompts, policies or learned behavior??

A persistent conversation object with channel-agnostic events turned out to be the right abstraction for cross-channel context, rather than a thread-per-channel stitched at read time. Took a few iterations to land on, and most of that work isn't visible from the outside, which is exactly the bar we set on the engineering side. A customer moving from WhatsApp to email shouldn't have to notice the platform at all.

Proud of the team for getting this out. The simulator does both: historical replay against new instruction versions for regression coverage, and synthetic generation for the long tail.
We argued on the engineering side about which to build first, and eventually decided neither one on its own makes the evaluation loop useful rather than theatrical. Both landing cleanly is the part I'm most invested in.

Response-quality grading on its own never catches the interesting failures. Action-sequence validation against an expected workflow, invariants on which tools get called for a given intent, custom policies beyond simple output checks; that's where the real agent bugs live.

Getting that into the harness as a proper API rather than a checkbox was the thing we kept pushing for on the QA side.

A defensible ROI estimate in 60 seconds shortens discovery by a full call when you're running against an incumbent.

Saw it during a live demo last month, which is not something calculator widgets usually pull off. Small feature, outsized impact, and a detail I kept advocating for from the GTM side. A proper end to end solution for once!

The conversation around AI agents shifted. It's no longer about how to build one. It's about how to manage thousands of them.

What it is: Gemini Enterprise Agent Platform is Google Cloud's unified environment to build, deploy, govern, and optimize AI agents across enterprise infrastructure.

Problem → Solution: Prototyping an agent is solved. The hard part is what comes after — assigning identity across a fleet, maintaining context across multi-day workflows, detecting anomalous behavior, and debugging failures in production. Vertex AI handled the model layer. It left the operational layer to custom engineering. Gemini Enterprise Agent Platform closes that gap with first-class tooling at every stage of the agent lifecycle.

Key features:

• Agent Studio: visual canvas for building, testing, and prompt comparison without leaving the development environment

• Agent Runtime: sub-second cold starts, agents that run autonomously for days

• Memory Bank with Memory Profiles: persistent, low-latency recall of user-specific context across sessions

• Agent Gateway: centralized security enforcement across all agents regardless of origin

• Agent Simulation and Evaluation: test against synthetic interactions, score against live traffic

• Model Garden: 200+ models including Gemini 3.1 Pro, Gemma 4, and third-party models including Claude

Benefits:

• Engineering teams ship production-grade agents without building custom governance infrastructure

• Security teams get a unified identity and audit layer across the entire agent fleet

• Operations teams get real-time observability and anomaly detection without manual log review

• Product teams get pre-built agent templates in Agent Garden covering invoice processing, financial analysis, and code modernization

Who it's for: Enterprise engineering teams and platform architects at organizations running or planning to run multi-agent systems on Google Cloud.

Six trillion tokens flow through ADK every month. This is the infrastructure Google built to make that scale governable.

P.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends

Hey Product Hunt! Johnny here from Fabric 👋

We're back! And this time we built something specifically for the developers and tinkerers in the community.

Today we're launching Fabric CLI 🖥️

Your entire Fabric library, right from your terminal. Search it, save to it, talk to it.
Never tab out again.

✦ What is it?
A free command-line tool that connects to your Fabric workspace. One line to install, no dependencies. You get:
• fabric save "revisit the auth retry logic"
• fabric search "mixture of experts diagram"
• fabric ask "summarize everything tagged with project-atlas"

AI search that finds what you meant in about 200ms. An AI agent that can summarize, tag, move files, create notes. All without leaving the shell.

✦ Why we built this

We kept hearing from developers: "I love Fabric but I live in my terminal." Fair enough.
But there's a bigger reason too. Coding agents like Claude Code, Codex, and Cursor are incredible, but they forget everything between sessions. They have no persistent memory.

With the CLI, any agent that can run a shell command can pull context from your Fabric library before starting a task and save what it learned when it's done. Your agents get smarter over time instead of starting from scratch.

✦ Quick recap of Fabric for the uninitiated:
🧠 An AI workspace that organizes itself around how you actually think
🤖 Frontier memory engine (more on this soon)
🔍 Semantic search across everything you've ever saved
📱 Apps on web, mobile, desktop, and now the terminal

Free to use. One line install. Give it a spin and let us know what you think!

We read every piece of feedback and it genuinely shapes what we build next.
Thank you for being part of this with us 🙏

I basically live in the terminal and kept alt-tabbing to the app to save things or look stuff up. now I don't. one line install, no deps, and it just works

Personally, very happy that we made this – super handy to save a quick thought without leaving the terminal!

This is basically how I use Fabric now. I just pipe things into fabric save without really thinking about it. The search is surprisingly fast!

Hey Product Hunt! 👋

I'm Aidan, one of the builders of @Docsio

Docsio turns your product into a fully branded docs site in minutes, editable by chatting with an AI agent.

We built this for small SaaS founders who need proper docs but don't want a janky DIY setup or a $300/mo bill.

If you're shipping a product and your docs are an afterthought (or don't exist yet), this is for you.

✍️ Here's how it works:

1. Paste your product URL, or drop in your own content (Notion, PRDs, specs, whatever).

2. We scrape your site, pull your brand (colors, logo, fonts, favicon), and build a complete docs site with real content. You watch it happen live.

3. Chat with the AI agent to edit anything. One click to publish with SSL and a custom domain.

Why you want Docsio:

• AI agent editor: Like Cursor or Claude Code, but for docs. Full file access, so if you can describe it, it builds it.

• Auto brand extraction: Your docs look like your product from the first preview. Zero CSS work.

• Live preview: Sandboxed dev server with hot reload and mobile preview. What you see is what ships.

• AI chat widget: Every published site gets an "Ask AI" trained on your docs.

• MCP server: Plug your docs into Cursor, Claude, and other AI coding tools.

• Isolated sandboxes: Your content is yours. Nothing shared, nothing trained on.

🎉 For the PH community: To celebrate the launch, use code PHLAUNCH50 to get 50% off a Pro subscription.

Paste your URL and show us the docs site it builds for you 🤎

P.S - Noticed Product Hunt's docs could use a little love, check out what it made with just the URL!

Within minutes we went from this:

to this:

Most doc sites price like this:

Whereas Docsio prices like this:

It's easy!

I'm Jameson, another one of the Cofounders of Docsio.

And we really couldn't be more excited to share this with the PH community today.

We built Docsio because we wanted something smarter than what was out there.

Would love to hear your feedback, answer any questions, and connect with anyone who's interested.

Drop a comment below, I'm here all day!

– Jameson

This is a good one ! But docs need to address so many pain points ! Does this start with a "Getting Started".

If you’ve ever:
**spent weeks building technical docs
**paid hundreds per month for simple tools
**fumbled an onboarding due to outdated docs
**or pulled engineers off valuable work and into documentation

this one’s for you.

Hey, congratulations on the launch 🎉

Really great tool - I just created a docs site for my startup, Fendemo.

But I’m running into an issue when trying to connect a custom domain. It shows this error:

Cannot add docs.fendemo.com since it’s already in use by one of your projects (pending verification).

It's been a thrill to build Outpost. We took our experience receiving 100B webhooks from over 7k different vendors and condensed these learnings into a straightforward API and product. It's the fastest and most scalable way to send webhooks.

Hey Product Hunt, and thanks to @fmerian for hunting!

We launched Outpost as an open-source project about a year ago to solve a problem we kept seeing: every platform that needs to send webhooks to their customers ends up building the same infrastructure from scratch: retries, tenant isolation, delivery logs, a portal for end users to manage their endpoints. It's a significant engineering investment for something that isn't anyone's core product.

Outpost was our answer to that: open source, self-hostable, and designed to be the last outbound webhook system you need to build.

Outpost doesn't just do webhooks. It supports Event Destinations, too: your customers can receive events via SQS, Pub/Sub, Kafka, RabbitMQ, and other brokers and queues, not just HTTP requests. This is something we've seen Stripe, Shopify, and Twilio all moving toward, and Outpost gives any platform the same capability out of the box.

We also now offer a managed version (Hookdeck Outpost) for teams that don't want to run the infrastructure themselves. It runs the exact same codebase (no proprietary fork). Pricing starts at $10 per million events, which makes it the most affordable managed option available.

Would love to hear from anyone who's currently maintaining a homegrown webhook delivery system. What are the biggest pain points? And if you've tried Outpost, what could we do better?

Hey Product Hunters 👋 I lead DX at Hookdeck and have been working across the Outpost SDKs, docs, and parts of the codebase.

When we conceived Event Destinations and started building Outpost as the first open-source implementation, we thought it would be interesting and hoped it would land. It's done far better than we expected. Since the beta, we've worked hard with the open-source community and customers, and one of the things we hadn't specifically planned for was a managed version, but the demand for it was clear enough that we built it. That's what's launching today alongside the open-source v1.0.

Outpost is a manifestation of Event Destinations. Platforms need to deliver events wherever their customers actually want them, and to offer a really solid webhook option. HTTP isn't going anywhere, but for many use cases, delivering directly to a queue, stream, or broker is a better fit. Outpost covers both, and it's the best outbound event infrastructure available right now. It's simple, and with that comes really solid DX and ease of integration.

Outpost also fills a clear gap on the agentic side. So many companies are building with agents, for agents, or integrating with agents, and event infrastructure is foundational to a lot of what they need. Agents need to be triggered by events, and agents need to emit events when things happen inside their workflows. On task completion, on state changes, and on handoff. That's what enables agent-to-agent communication, event-driven agent initialization, and workflows where one agent's output triggers another agent's run.

Always looking for feedback on the product in general and on DX in particular.

Congrats on the launch!

We've been using Outpost since the beginning of the year while in beta and it's been great! Y'all are super responsive with feedback or issues, pricing is more than fair and generous free tier, and the UI is really polished too. Really enjoyed working with you guys and we're a happy customer :)

Wow, congrats on the launch, team! This is such a real infrastructure problem, and Outpost feels like a very clean, thoughtful answer to it. Love the open-source plus managed approach too. Excited to see this launch.

Congrats team for the GA.

We've been running Nuntly's webhook delivery on self-hosted Outpost for months. Best decision we made: ripping out our in-house implementation.

Not going back.

Hi everyone!

This is a dense model, not an MoE, and that matters.

Dense models often show unusually strong intelligence density for their size, and Qwen3.6-27B is a very good example of that. At 27B, it already pushes past Qwen3.6-35B-A3B on a number of key coding and reasoning tasks, and more importantly, it beats the previous open-source flagship Qwen3.5-397B-A17B across all major coding benchmarks. That is a pretty serious result for a dense checkpoint at this scale.

And 27B is also just a very sweet open-source size. It is not so large that normal users or small teams are locked out of deployment, but it is not small either — it still leaves a lot of headroom for real capability.

In the Qwen3.6 era, this has a very real chance of becoming their most popular open dense model.

we're almost there. if that long term coding gets bumped up and reaches opus 4.5 levels we're looking at something serious. i reckon in about 6-9 months they are there or beyond that and approaching opus 4.6 levels. at that rate, for coding, running that locally on your hardware, it's hard to justify picking a frontier cloud model.

Hey Hunters, I am excited to hunt 🚀

Today I found something really interesting from ChatGPT — Workspace Agents.

In simple words, these are like AI teammates that can actually handle work for you. Not just small tasks, but proper workflows.

You can create an agent once, tell it what to do, and it can:
– work across tools like Slack and Linear
– keep track of progress
– and move things forward without you checking again and again

What I liked most is that you can set it up based on how your team already works. So it doesn’t feel random—it follows your style.

You can use it for things like checking leads, managing feedback, making reports, or even doing research.

It’s still in preview, but honestly this feels like a big step towards real AI teammates.

Curious—what work would you give to an agent first? 👀

I've been using IFTTT for 15+ years, yes, I was automating things before it was cool. As a self-proclaimed automation freak, this MCP server news basically made my whole week. The fact that I can now integrate Claude directly into my IFTTT workflows is the crossover episode I never knew I needed. It's like peanut butter finally meeting chocolate.

Thank you for continuing to make the nerds of the world deeply, unreasonably happy, we see you, and we appreciate you! 🚀

I like how you guys embraced markdown!

This will forever go very well with Lovable type applications.

love that you can request missing brands. the format makes so much sense - been frustrated with trying to describe design systems to AI in prompts when there should just be a standard way to encode this stuff. shipping this as markdown files instead of some proprietary format was smart.

this is brilliant. we've been using Cursor daily and always end up with that generic AI beige you mentioned. having a DESIGN.md file that actually captures brand systems in a format the AI can parse feels like the missing piece. curious how detailed you can get with component behavior - like does it understand hover states and micro-interactions, or mainly static properties?

Hi everyone!

Xiaomi MiMo ships new models at a pace that’s honestly astonishing. MiMo-V2.5-Pro is their strongest yet — big jumps in agentic coding, long-horizon tasks, and real-world reliability. The lighter V2.5 brings native multimodal (image/audio/video) on top of that while staying very token-efficient.

MiMo is clearly accelerating to become one of the fastest-moving players in the frontier tier.

Right now you can try it in MiMo Studio and use the whole series through their Token Plan, and the models will be open-sourced soon!

Questions: why is XAI so overweighted? Are you planing to add more companies to the portfolio and spread the risks a little bit, so that it is priced more as industry average and if one CEO screws something, one company doesn't bring down the performance of whole portfolio?

Any idea if this is only for US residents or can investors be from any geo?