PH热榜 | 2026-04-16

I'm co-hunting the Claude Code desktop redesign today with my PH friend @fmerian, and it's one I've been watching closely.

The problem: Agentic coding doesn't look like "one prompt, one answer" anymore. Developers are running refactors, bug fixes, and test passes across multiple repos simultaneously. The old desktop app wasn't built for that. Managing parallel sessions meant juggling terminals, editors, and context separately.

The solution: Anthropic rebuilt the Claude Code desktop app from the ground up for parallel agent workflows.

Here's what's new:

• Session sidebar — View & filter all sessions; auto-archive on PR close

• Side chat (⌘ + ;) — Ask questions without disrupting main flow

• Terminal + editor — Review, edit, and ship in one place

• Drag & drop layout — Customize your workspace grid

• 3 view modes — Verbose, Normal, or Summary

Especially useful for engineers already running Claude Code at volume who are hitting the ceiling of managing multiple sessions manually.

Update your app or download. It requires a Pro, Max, Team, or Enterprise plan.

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

the summary view mode is the most underrated part of this.  been running claude code for months — verbose output is the constant friction. you end up reading claude's narration instead of the diff. having that as a first-class toggle at the session level, not a system prompt hack, changes the actual workflow.

Using this for MentionFox and FoxAPIs right now and I can say it is a bit unstable, but usable. I have Claude managing it for Claude Code automation. I can say that you have to watch Claude code when automated. It will tell you things are done, but it didn't commit. Buttons that do nothing, pages not JSON'd together...

I've been using primarily Claude Code CLI so this will be very interesting to try out later this week. Having everything visible in one window sounds like a treat. I might find I needed this more than I could have anticipated, as someone that's used to the CLI and Linux Terminal I look forward to exploring this. Thank you

useful stuff ! i like it , what programing language is it native in?

awesome!!

Will there be a Linux version? 🤞

parallel agents in one workspace is the part i'm most curious about — how do you handle merge conflicts when two sessions touch the same file? does it surface them or just let the last write win?

Is this a blog post title or a product launch? who cares?

This is interesting, can I use it on the MAC? Is there a display option for UI config instead of CLI?

Just upgraded @Claude Code Desktop App Redesigned v1.3036.0. If you set any

@Claude by Anthropic env vars for terminal use (API keys, model overrides, custom base URLs), the Desktop app will pick them up and break with cryptic 400 errors.

Fix: move terminal-only vars to .zshrc (interactive shells only) and clear the macOS launchd cache:

launchctl setenv YOUR_VAR_NAME ""  

Then Cmd+Q and relaunch.

Fixed the env bug in 15 minutes. Upgraded to Opus 4.7. Back to building.

It's launch week at @Resend - a week of announcing new features every day.

So far, the team announced Automations, launched an AI Email Editor, and just released their new CLI.

S/O to @zenorocha and team - Looking forward to what's next!

I love resend and have been a user for quite a while now! Btw could resend eventually auto optimize or improve emails based on open rates or spam signals using ai???

nice to see a CLI that treats AI agents as a first-class user. curious what the skills format looks like — is it a standard spec or something resend-specific?

We live in a terminal all day anyway 😅. It just makes perfect sense to move more parts of the old UI to it and simply work faster!

@zenorocha the agent Skills feature is interesting — are these MCP-compatible or a custom spec? I'm thinking about agent pipelines where email is a side-effect trigger, and protocol compatibility matters a lot.

awesome! i love / evangelize resend as it's a. insanely easy to spin up b. works well and c. is very affordable when trying a lot of small projects.

I often setup some kind of "one off" email sender in my apps where i will have some panel that just lets me write / send emails to users manually (as opposed to automated campaigns, flows, etc). i am excited to give the cli a twirl.

webhook listening from terminal is clever for development. beats spinning up ngrok every time you want to test email events. how does it handle multiple webhook endpoints if you're working on different projects?

React Email support in the CLI is smart. debugging email templates locally before sending has always been painful. does this let you preview the rendered output or just validate the components?

the AI agent skills feature is interesting - what kind of email workflows are you seeing agents handle best? we've been automating a lot of our healthcare product notifications and curious how this compares to just hitting the API directly

Love the focus on dev experience + deliverability.
Feels like that combo is what most tools still don’t fully get right.

Hi Product Hunt — I’m Heshan, founder of X-Pilot.

After leaving Baidu Apollo, I built 3 edtech companies (1M+ users total) and kept seeing the same issue: the people with the deepest knowledge are often the least equipped to turn it into video. Hand a professor a video editor and everything slows down. I started calling it the “Expert Paradox.”

X-Pilot is our attempt to solve it: upload a document, and X-Pilot generates an accurate, multi-module video course—complete with a syllabus, learning objectives, and animated visuals (diagrams, rendered formulas, code walkthroughs) you can publish.

A key difference vs. HeyGen/Synthesia: those are talking-head/avatar script readers. X-Pilot focuses on knowledge visualization. Every visual is rendered programmatically via Remotion in isolated sandboxes—deterministic code, not generative visuals—so if your doc says 2+2=4, the video shows 2+2=4.

Free to start (no credit card).

I’d love feedback from anyone who’s tried to turn a document into a course: what broke for you—structuring, visuals, editing time, accuracy, or distribution?

Hey Product Hunt 👋 I’m one of the devs behind X-Pilot.

We built this because we kept hitting the same issue: AI video tools look great, but they hallucinate—especially on code, charts, and formulas. For educators and trainers, that’s a dealbreaker.

So we went a different direction. X-Pilot takes your PDF / PPT / Markdown and renders everything deterministically in a sandbox (via Remotion), instead of “imagining” the video.

That means what you write is exactly what shows up—no hallucinations, especially for technical content.

It's been a massive engineering challenge to build this rendering engine from scratch, but seeing it save course creators hours of manual editing has been incredibly rewarding for our team.

We’ve added free credits so you can try the real workflow. I’ll be around in the comments—would love feedback, ideas, or any questions.

AI video tools usually mess up charts. Does this fully avoid that?

Curious what the actual workflow looks like for a non-technical creator. Upload a doc — and then what? How many decisions do I need to make before I have something I'd actually want to publish?

The "Expert Paradox" is real — I see it constantly in finance. The best modelers often produce the worst training materials because they skip steps they've internalized. X-Pilot solving this for video is compelling. I teach Excel for financial modelling on Udemy and structuring course content that works for both beginners and intermediate practitioners is genuinely hard. A tool that can take a document and render accurate multi-module content around it would be a game changer for technical finance education. Congrats on the launch!

can I use it with product documentation to create help videos for my product?

Just found the free tier – 3 minutes/month, no credit card. Uploaded a test PDF with some code snippets. The "zero hallucination" claim held up – my diagrams rendered exactly as written.

Quick question: Does the free tier include natural language editing (like "shorten the intro"), or is that locked behind paid plans?

Asking because typing edits is way faster than timeline dragging. Overall, a promising tool. Going to test more.

In regulated industries, "close enough" isn't acceptable. X-Pilot's deterministic rendering means every policy explainer video we produce is audit-ready. When a regulation changes, we update the source document and regenerate — no re-shoots, no version control nightmares. Our legal team reviewed the output and signed off. That's never happened with an AI tool before.

Congrats On the launch

As a physics professor, I've always struggled to turn my LaTeX-heavy lecture notes into video. X-Pilot nails the formula rendering — every equation is programmatically rendered, not some blurry AI-generated image. I converted a 40-page quantum mechanics PDF into a 6-chapter video course in under an hour. My students' exam scores went up. This is what "accuracy-first" actually looks like.

Does it support importing multilingual documents? For example, if I upload a Chinese PDF, can it directly output a video with English audio?

What impressed me most is the pedagogical intelligence. X-Pilot doesn't just convert text to video — it structures content using Bloom's Taxonomy principles with proper scaffolding. The AI Syllabus Generator alone saved me 8 hours on my last course design. For anyone in instructional design, the learning objective mapping feature is worth the switch.

I'm a bit curious about the generation speed and cost. If it's stable, there should be a market for educational content.

When you mention 'knowledge visualization,' do you mean preset templates or support for custom styles? If I want to make the video in a minimalist style or with a brand color scheme, can I adjust that?

remotion under the hood is interesting… do users need to touch code?

Hi Product Hunt — I’m a developer at X-Pilot.

I’ve seen firsthand how much "dormant knowledge" goes to waste simply because video production is too slow and technically demanding for true experts.

X-Pilot was born to bridge that gap. We help you turn static documents into structured, animated video courses in minutes. But we didn't stop at just "making a video."

Why X-Pilot stands out:

Knowledge Visualization, not just Talking Heads: Unlike avatar-based tools, we prioritize precision. Complex diagrams, code snippets, and formulas are rendered programmatically, ensuring your content is sharp, accurate, and professional.

From Raw Docs to Finished Assets: Simply upload your materials. Our engine handles the structural planning, sequential chapters, and complex animations, transforming raw input into a polished, logical flow.

Natural Language Editing: Need to swap a chart, change a visual style, or rewrite a scene? Just use natural language. Our AI editor handles the heavy lifting, replacing tedious timeline scrubbing with prompt-based iterations.

We’re thrilled to be here on Product Hunt to get your feedback.

Knowledge visualization > Talking heads. The Expert Paradox is real, and this solves the biggest friction.

Generating a syllabus + rendered diagrams from just a doc is a huge time-saver. Are the visual templates customizable via CSS or code?

As someone who's spent way too many hours wrestling with video editing tools just to turn a technical document into a decent course video, X-Pilot is a breath of fresh air.

What immediately sets it apart: zero hallucinations. In education and training, accuracy isn't optional — it's everything. The fact that every visual is rendered programmatically via Remotion in isolated sandboxes means my formulas, code blocks, and diagrams come out exactly as they should. No AI "reinterpretation," no subtle errors that erode trust with learners.

I uploaded a 45-page technical PDF and had a structured, narrated video course in under 20 minute— something that would've taken me an entire weekend with traditional tools. The auto-chaptering is smart, the Visual Motion Boxes make concepts genuinely engaging, and the natural language editor ("make the intro shorter", "swap the chart in scene 3") feels like magic compared to dragging clips on a timeline.

A few things I particularly appreciate:

Content fidelity — E = mc² stays E = mc², not some AI-hallucinated approximation

The NL editor — 36x faster than traditional editing is not an exaggeration in my experience

SCORM export — huge for anyone deploying to an LMS

Multilingual translation — with voiceover sync preserved

If you're a course creator, educator, or L&D professional who needs accurate training content at scale, this is the tool you've been waiting for. The free tier gives you enough to test the full workflow — I'd strongly recommend giving it a shot.

Congrats to the team on the launch! 🚀

Congrats! Does it support file input like image and .docx files, besides text?

Skills in Chrome is a new feature that turns your best AI prompts into reusable, one-click workflows right inside the browser.

The problem: repeating the same prompts across tabs and tasks is tedious and inefficient.

The solution: save prompts as “Skills” and instantly run them across pages or multiple tabs.

What makes it interesting is the ability to remix, reuse, and execute prompts contextually on live webpages, not just in chat.

Key features:

• Save prompts as reusable Skills from chat history

• Run Skills across current or multiple tabs

• Built-in library of ready-to-use workflows

• Edit and customize anytime

• Privacy + confirmation safeguards for actions

Benefits: faster workflows, less repetition, and more powerful AI-assisted browsing.

Great for: researchers, shoppers, productivity geeks, and anyone using AI daily.

Use cases: comparing products, summarizing documents, analyzing info, and more.

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

I love this idea, I've been manually reusing prompts forever and this feels like a huge save of time for me already.

Is there a way a website can create custom skills for interacting with there website?

I love this, I have a bunch or docs with prompts, when I have a really good one that gave me the results I want, those are gold. It took me a long time to get good at it and still have a ways to go. This would make it easier, don't need to g digging through files.

Interesting launch. I think this will solve a lot of workflow problems for Chrome Users.

Solid focus on control and privacy , being able to pause and redact is a big win for real world meetings.

Hey Product Hunt! 👋

Today we're shipping one of the features our users have been asking for most: Fellow goes fully mobile. Recording, transcription, AI notes – all from your phone.

Fellow was built on a simple belief: your meetings are some of the most sensitive conversations your company has. They deserve more than a bot with loose data practices. We built the most secure AI meeting agent — SOC 2 Type II, zero data retention, botless by default. We built that on desktop. Now it's also in your pocket.

Here's what's new:

🔴 Botless mobile recording: Hit the red button, and Fellow captures your meeting through your device's mic. 

🤖 Ask Fellow on mobile: Our AI chief of staff is fully functional on the app. Ask questions about past meetings, generate follow-up emails, get instant insights from your phone.

📅 Full calendar & agenda management: View upcoming meetings, edit agendas, toggle recording on or off, and show up prepared no matter where you are.

✅ Action items on the go: Accept, dismiss, or create action items between back-to-back calls. 

🔍 Your full recording library, in your pocket: Every transcript, AI summary, and recap, accessible anywhere.

Available now on iOS (v1.6.0+). Android recording is coming very soon.

We built Fellow to be the most secure AI meeting agent — and now it goes wherever you do. 

Would love to hear what the PH community thinks. 🙏

Definitely saves a lot of trouble. Will it alert those on the phone that the meeting is being recorded? A lot of services out there require permission for this sort of thing.

I use this all the time, biggest use case is getting all my meetings in one place and figuring out what the hell is going on!

Tools like Fellow are must haves to work effectively with AI and Agents - where context is everything. Congrats on the launch of iOS app @aydin_mirzaee & team!

Okay, let me get this straight. You're telling me I can do in-person meetings now and just tap a button and record? This is going to be a game-changer for me. I've been waiting for this for a while. Can't wait to test it out. I'm on this.

Nice balance between AI automation and control (pause/redact is key).

Feels like trust is a big differentiator here vs other note takers.

Hey Product Hunt 👋 I’m Glenn, one of the people behind stagewise.

We built stagewise because most coding agents are still too blind to the actual product - and the easy to use ones are also pretty closed off! They can work with code, but they usually can’t really see the app, inspect live state, or understand what’s happening on screen without a lot of manual context or manual harnessing that takes time and is an additional burden for users.

stagewise changes that.

It’s an open-source coding agent with its own browser environment. It can see what you see, read the DOM and console, inspect live app state, and edit the code behind what it’s looking at.

We also wanted the setup to stay flexible: bring your own API key, use any model/provider you want, run local models, and avoid lock-in. Our product is fully open-source and we want to keep it that way - pushing the boundaries of what's possible for agents and making it accessible to everyone.

If you’re building frontend products, websites, or interactive apps, our goal is simple: With stagewise, you should need less explaining, less context switching, and faster iteration.

Happy to answer questions, hear criticism, and learn what workflows you’d want to use this for.

- Glenn

Let's go! This is genuinely the best agent/ desktop app one can currently use for working on the web, no doubt

How stable it feels in complex, component-heavy React apps though?

I find stagewise interesting because I often struggle with reproducing bugs locally. If it can truly read console output and DOM state while linking that directly to code edits, I can see myself using it daily.

Stagewise has been incredible to work with for the past weeks. I'm really missing it whenever I don't use it, especially since it makes so many things so easy, like taking screenshots in both dark and light mode for my docs, or copying some colors from another website and testing around with them.

Amazing product and team! :fire: We used @stagewise to build our new rebulk.com website and it absolutely killed it.

@glenntoews the browser environment model is interesting — does stagewise see the DOM as raw HTML or a semantic tree? I'm wondering how it stays coherent when the component owning a UI element is 3-4 files away from the markup.

Gemini 3.1 Flash TTS is Google's new text-to-speech model, now available in preview via the Gemini API, Google AI Studio, and Vertex AI.

The problem:

TTS APIs have always treated voice as a static output.

You pick a voice, set a speed, and the model delivers a flat read.

Getting expressiveness meant engineering workarounds or accepting robotic delivery.

The solution:

Gemini 3.1 Flash TTS introduces audio tags natural language commands embedded directly in the text input to control tone, pacing, accent, and expression mid-sentence.

You can define scene context, cast multiple speakers with unique voice profiles, and export the full configuration as API code for consistent reuse across projects.

What stands out:

🎙 Inline audio tags mean you can shift tone, pacing, and delivery mid-sentence without re-prompting

🗣 Native multi-speaker dialogue means you can cast and direct multiple characters in a single API call

🌍 70+ language support with per-locale accent control means you can localise expressive speech without a separate pipeline

📤 Exportable voice config means your characters and delivery style stay consistent across every projec

🔒 SynthID watermarking means every output is attributable as AI-generated out of the box

Who it's for:

developers and product teams building voice agents, AI dubbing tools, interactive storytelling apps, and multilingual content platforms that need expressive, controllable speech at scale.

I did the tests, oh my god, it turned out amazing.

the inline audio tags unlock something specific for interactive web apps — not just narration, but contextual feedback. building with voice input, you always want the confirmation  to sound different from the question, which meant separate prompts or post-processing hacks. being able to embed that context inline changes the design space for conversational interfaces.

Hey Product Hunt!
I'm Nikola, founder of ClayHog, and I’m excited to finally share what we’ve been building 🚀

If you’ve been following the shift to AI search, you’ve probably noticed something strange: your brand shows up in ChatGPT, Perplexity, or Gemini… but you have no idea when, why, or how often.

That’s the problem we’re solving with ClayHog.

ClayHog helps you understand how your brand appears in AI-generated answers and gives you the tools to improve it.

Think of it as visibility tracking + prompt intelligence + content creation for the AI-first web.

Instead of guessing, you can track prompts, discover opportunities, monitor citations, and create content that increases your chances of being mentioned.

What makes ClayHog stand out:

• Prompt Tracking (Scouts)
  Track the exact prompts your audience is searching for and see how your brand performs across ChatGPT, Perplexity, and Gemini.

• AI Keyword / Prompt Discovery
  Find high-impact prompts and questions people ask AI tools, not just traditional search keywords.

• AI Visibility Dashboard
  Understand your presence with insights like visibility, sentiment, and competitor comparisons, all in one place.

• Citation Tracking
  See where AI tools pull information from, which sources mention you, and where you’re missing.

• Content Opportunities & Creation
  Discover what content AI prefers for your niche and generate ideas or drafts that increase your chances of being cited.

• AI Crawler Logs
  Monitor visits from GPTBot, ClaudeBot, PerplexityBot, and more, and understand how AI crawlers interact with your site.

• GEO Audit
  Run a quick AI search readiness report and see how well your site is prepared for AI-driven discovery.

…and much more, from Reddit & YouTube monitoring to domain signals, brand-guided content, and shareable reports.

ClayHog started as a simple question: “How do we know what AI says about us?”
It quickly became clear that this is a completely new layer of visibility, one that traditional SEO tools don’t cover.

Now it’s not just about tracking, it’s about taking action.

We’re just getting started, and your feedback means everything to us.

If you're thinking about AI visibility, GEO, or just curious how your brand shows up in tools like ChatGPT, give ClayHog a try.

Drop a comment, share your thoughts, or DM me , I’d love to hear what you think 🙌

Cheers,
Nikola 🦔

Not sure if this is just happening on my end, but I ran into an issue with the signup flow.

I signed up using my Google account and completed the form with my details and brand website. However, when I clicked “Next,” nothing happened. I tried again across different browsers, but the issue persisted.

Wanted to flag this in case others are experiencing the same, happy to provide more details if it helps debug.

Different AI tools often have very different takes on the same brand. When I see ChatGPT mentions me differently than Perplexity, how do I figure out what's actually driving that gap?

This is truly an underrated product/service. One of my biggest anxieties is about how I can ensure I get maximum brand exposure. Your tool looks like a great start alongside Google search console.

How long should a brand have operated for in order for ClayHog to gather sufficient information for its analysis?

Great product Nikola 👍 I'd like to see how it tracks our position overtime and makes relevant content suggestions. I wasn't able to produce suggested content on the free trial but the outline it suggested made sense which makes it a very useful feature. Is the content AI friendly and do you provide the associated meta data for LLMs?

Cool Nik! Sounds awesome. Just guessing if it works to analyze competitors presence? I mean, if I can check what AI tells about my competitors

Given that tools like ChatGPT and Perplexity generate non deterministic answers, how do you ensure tracking consistency?

That's a very cool product. I've already added my product to track it. The Reddit section has been especially useful for me at my current stage and already helped me to find places where I can show up and share what I'm building. Prompt tracking looks really useful too. Like it, following.

Any aspiration to create something similar for personal brands that are present on social media? :)

Super interesting launch 👀

What it is: Gemini CLI subagents — a system that lets your main AI agent delegate tasks to specialized “expert” agents.

Problem → Solution: Complex workflows overload a single agent’s context and slow things down. Subagents solve this by splitting work into isolated, task-specific agents that return clean, summarized outputs.

What makes it different:
Instead of one overloaded AI, you get a coordinated team of agents working in parallel — each with its own tools, context, and instructions.

Key features:

• Parallel execution of multiple subagents

• Isolated context windows (no context pollution)

• Custom subagents via simple Markdown configs

• Built-in agents for codebase analysis, CLI help, and general tasks

• Easy delegation using @agent syntax

Benefits:

• Faster execution on complex tasks

• Cleaner context → better outputs

• Scalable workflows for dev teams

Who it’s for:
Developers, AI builders, and teams working on large codebases or multi-step workflows.

Use cases:

• Codebase exploration & debugging

• Parallel research or analysis tasks

• Automated workflows with custom agents

• Enforcing coding standards across projects

This feels like a shift from “using AI” → “managing AI teams.” 🚀

Great! congrats on the launch :)

isolated context per subagent is the right call — curious how scoped permissions work in practice. can the main agent grant a subagent temporary shell access for a single task, or is it declared upfront per subagent definition?

Hey PH 👋

Picture this: your first investor just said yes. You're riding high. This is the moment you've been grinding toward for months.

Then someone says, "just send over the SAFE agreement."

And just like that, the high evaporates.

You open a new tab. Then another. Then five more. You're Googling "YC SAFE template," watching a YouTube video from 2019, texting a founder friend at 11pm asking if they used a lawyer for this. You're not sure if you need the pre-money or post-money version. You're not sure if you're supposed to fill in the valuation cap yourself or leave it blank. You're not sure if you need a lawyer to review it (and if you do, how much that's going to cost).

You get the document sorted. Now you need DocuSign. You sign up for a free trial, realize it requires a credit card, and decide to just figure it out later. You email the PDF to the investor. They ask a question. You answer. They ask another. A week passes.

The SAFE gets signed. You exhale.

Then you remember: you still have to update your cap table.

We built Mantle SAFEs because that experience is completely unnecessary, but it's happening to thousands of first-time founders right now, at exactly the moment they should be celebrating.

Here's what changes with Mantle:
→ YC SAFE templates built in
→ E-signatures handled inside Mantle, no DocuSign subscription needed
→ Cap table updates automatically the moment the SAFE is signed
→ Full document history and audit trail in one place
→ Free to use

From "want to invest?" to a signed SAFE in the same afternoon.

We'd love your feedback as we keep building. Drop your questions or thoughts below 👇

Hi everyone!

@OpenAI ’s updated Agents SDK adds two big pieces for production agents: a model-native harness for long-horizon work across files and tools, and native sandbox execution (including @cloudflare, @Modal, @Vercel and @E2B) so agents can inspect files, run commands, edit code, and keep working safely in controlled environments.

The production agent infrastructure race is getting very real!

I'm particularly interested in how the harness handles long-horizon agent tasks; I often struggle with state tracking.

native sandbox execution across E2B/Modal/Daytona/Vercel is a nice touch — does the SDK abstract away the provider differences or do you still pick one explicitly per run?

@zaczuo curious about the sandbox provider abstraction — is there a standard interface so you can swap between E2B, Modal, Daytona, and Vercel, or does choosing one lock you into specific execution semantics? That portability question matters a lot for teams already invested in one runtime.

finally, a harness that doesn't feel like an afterthought. been using Cursor and Claude Code daily and the safety guardrails are always the weakest link. love that you're tackling file inspection and command execution head-on instead of just wrapping the API calls.

I like how you framed this. I’ve struggled with passive sites too and this feels like a genuinely practical fix.

I’ve always felt most websites just sit there doing nothing. What you’re building sounds like it actually engages users instead of waiting.

Oh, I love the concept - seems really helpful. I wonder if I missed the info, but do you only use the data from the website? I hope you have a great launch.

Interesting take on conversational website experiences. We tried adding chat widgets to our B2B services site a couple years ago and the biggest challenge wasn't the tech — it was training the bot to handle the 80% of visitors who aren't ready to buy yet without being pushy. How are you thinking about the balance between engagement and not annoying visitors who are just browsing? Also curious if you've seen meaningful differences in conversion lift between B2B services sites vs. e-commerce.

I asked voice about pricing it says someting different than the actual pricing page worth noting looks cool though

Interesting concept! turning a static website into an interactive sales layer instead of a chatbot.

Curious about real-world conversion impact versus existing chat solutions.

Hi, Sohazur, another great product. I already checked the website out and created an account for myself. I'm looking forward for you to add German as another language because that's what I need. I'm pretty sure that some of my clients could use that tool as well. Cheers!

Very excited to launch this voice sales agent for your website, and what a timing that Product Hunt also collaborated with Wispr Flow and the voice dictation, because voice is the future of interface. And still, most of the websites don't have a voice. It just sits idle, and users are passively reading it. Foyer solved that problem. That's why we built it: not only to give the website a voice but also to make it interactive and turn every visitor into a sale. Foyer is free to use, and for the paid plan, we're giving 30% off to anyone who comes from Product Hunt. Go break it!

I’m wondering how flexible the spending limits are per agent. If I can control budgets tightly, I’d feel much more comfortable scaling autonomous actions.

Wow!!!! Love it...budget per day per card possible? Would be a killer feature.

This looks awesome. love the approach of giving agents more control over how they spend and how much they spend and to use this from the ai i already use! Going to give it a try!

This is really important. I had an open Flow agent set up with my current debit card, and I just worked a bill of $34 within one session. It is really needed to be aware of setting limits and doing all these things. I think a prepaid card would really help because, in that way, we can say open unknown expenses and all those things. This is really good. I really liked the idea. Congratulations on the launch.

Hey Product Hunt! We built AgentCard because we kept running into the same problem: AI agents are getting incredibly capable, but they can't pay for anything.

We wanted our agents to book flights, order groceries, and buy SaaS subscriptions — but giving them a real credit card felt terrifying. So we built Agent Card.

Here's how it works:
1. Fund your wallet via tripe
2. Issue a one-time-use virtual Visa card for our agent
3. The card auto-cancels after a single payment — no runaway charges

It works with Claude Code ChatGPT, and any MCP-compatible agent through our CLI or API. We'd love your feedback — happy to answer ny questions!

Hey PH! Matt here, cofounder/CEO of Astropad.

Since AI agents have put machines like the Mac Mini back in the spotlight, the need for reliable remote access has never been higher. But most remote desktop tools are still built for IT support, not devs running agents.

The problems we kept hitting:

• When an agent hits a snag, you won't know until you're back at your desk.

• You can't check in from your phone, nudge a stuck task, or restart something remotely.

That's what we built Workbench for. I use it daily to keep an eye on my OpenClaw instances from my phone.

A few things that make Workbench great for checking in on agents:

• We added voice input to dictate prompts and commands directly to your Mac.

• It's Apple-native with apps built specifically for Mac, iPad, and iPhone.

• And it's powered by LIQUID, our proprietary streaming tech we've been refining for a decade across our products Astropad Studio and Luna Display.

I'd love to get your feedback on Workbench — it's free to use for 20 min/ day (or there's an option to upgrade to unlimited).

I'll be in the comments all day answering questions!

Got an ad for this on X a few days ago and have been using this to control my mac mini at home - no openclaw needed! Nice work gang.

I've been using it for a few days, and it works really well!

Hey everyone! I'm Jonathan, founder of Avec.

I'm building Avec because email had become the single biggest tax on my attention. Not because of volume, but because of how much low-value decision-making it forces on you: what to read, what to skip, what to reply to, what to ignore. Every email is a to-do list someone else wrote for you.

Avec uses AI to give you that control back. It prioritizes what actually matters, handles the noise, and lets you focus on the messages worth your time with intuitive, delightful AI-backed features. My favorite is the ability to quickly respond to emails with voice, in mere seconds.

We've been in TestFlight for months with a group of early users who helped shape the product significantly. We just launched on the App Store, and I'd love for you to try it.

Happy to answer any questions about the product, the approach, or what's coming next.

You should know how many spam emails I get on Gmail, and sometimes I ignore the really important ones because of those spam emails. I am trying to manage them manually, but it takes a lot of time. I hope this also has a feature where the agent will automatically unroll me from all the unnecessary subscriptions. If that's there, then yeah, it's really worth it. Thank you!

i dont really care much about the ai features but ui is so clean and unified inbox alone makes it so precious to me

Seems really helpful, maybe not for me (I love going through my emails) but I know many people struggle with the clutter. I wonder though - how many languages are supported for the voice note option? Is it English only for now?

I look forward to emails just to try the voice reply/polish feature. Looking forward to your launch for outlook emails!

I like the idea of managing things with a swipe (my own app does this) but for email it's a bit more touchy. I do think that if you have high volume of email that could benefit from quick replies, this sounds like a great solution. It's less of a fit for my use case.

Tinder for email is such a good framing because you are optimizing for decision.

For voice note to mail, do you keep personality/tone of the speaker? is there filtering of words or does it stay a little rough?

Why so focused release? iOS only and US only... iOS, ok, understand because code. But what the difference for Gmail to other countries? 😢

applied to three jobs over a long weekend once. heard back from one by tuesday — not fit, just timing. the sunday night problem is real.

The 'always-on, zero-bias' approach here is a genuine step forward for hiring. The fact that a customer screened 2,000 candidates in a single weekend before launch says everything, this isn't vapor ware, it's a working product solving a real problem. The combination of lifelike avatars for authentic conversations, KYC-grade identity verification, and real-time proctoring makes OnScreen stand out from the usual batch of AI hiring tools. Congrats on the launch, team HackerEarth! 🚀

The "always-on, zero-bias" approach here is a genuine step forward for hiring. The fact that a customer screened 2,000 candidates in a single weekend before launch says everything — this isn't vaporware, it's a working product solving a real problem. The combination of lifelike avatars for authentic conversations, KYC-grade identity verification, and real-time proctoring makes OnScreen stand out from the usual batch of AI hiring tools. Congrats on the launch, team HackerEarth! 🚀

For a recruiter who is now seeing the number of applicants double/triple and each resume feels like it has been tailored to the job description, this is a game changer. The fact that it does not get tired and can screen a candidate even on a Friday night without any prejudices means every candidate gets the same experience. I think this will significantly change the way resume screening is done today.

Interesting objective hiring in coding interviews is still a huge problem.
How do you balance standardization with letting candidates show creativity?

Congrats on the launch team. This is really great

Interesting, does it only do English interviews or can it take the interview in other languages as well?

Congrats on the launch, @zahra_khan_hackerearth and team! This is fantastic. Really curious if there's a human-in-the-loop at any stage to handle edge cases?

how does it handle candidate stress? high-stakes assessment triggers nervous system dysregulation - performance in fight-or-flight looks nothing like baseline. curious if the scoring model accounts for this.

I love the idea so much, I might have to quit using Wispr Flow :P

Difference between this and mechvibes?

Hey product hunt! 🦖

I built Chinilla because I wanted a tool to help me understand how systems work. I tried learning through books and guides, but they're jargon heavy and still didn't give me the intuition of actually seeing and understanding systems visually.

So I built a flight simulator for it. You design a system visually, hit play, and watch traffic flow through it in real time. When your database chokes or your queue overflows, you see it happen. You fix it, run it again, and build the kind of intuition you can only get by watching things break.

What you can do:

• Drag and drop 7 universal building blocks

• Wire components together and define simulatable metrics and behaviors

• 12 programmable behaviors: queue, retry, filter, batch, rate limit, circuit breaker, and more

• Hit play to simulate and watch real packets flow through your architecture

• See bottlenecks, drops, and queue pressure as they happen

• Scrub a timeline to inspect any moment frame by frame

• Describe a system in plain english or paste code or text (frameworks, papers, etc), Chinilla AI maps it out on canvas

• Collapse groups, enter subsystems, explore at any level

• Find weaknesses with stability analysis, Monte Carlo with SLO targets, and stress testing

• Export to PNG, GIF, SVG, Mermaid, Python code, or a markdown spec

• Publish a live interactive link with a MD embed for papers or repos

• 16 templates to learn by doing (ML pipeline, coffee shop, chat app, rate limiter, and more)
  

It's free to use in your browser. The demo takes about 60 seconds and doesn't need an account: chinilla.com/demo

Free account gets you 5 cloud projects, full simulation, all export formats. No credit card.

Poured a lot of 💖 and time into this project. I'd love to hear from y'all what you think!

What's next:

• Team collaboration (shared canvases, real-time cursors)

• Fully fleshed out Duolingo style interactive lessons that teach system design step by step

• Tighter simulate-fix loop (inline suggestions when things break)

• Integrations (import from Terraform, CloudFormation, Docker Compose)

• Better UX polish based on your feedback (seriously, tell me what's rough)

Launch special: LAUNCH50 for 50% off Pro monthly (until April 30)!!

Thanks for reading 😃

To say I'm surprised is to say nothing! Cool idea!

P.s. I'm the only one who didn't get this link to open

This is a sharp idea — most teams only discover their system design flaws after deploying to production and watching things break under real load. We went through exactly this scaling our internal automation infrastructure. The simulation-first approach could save a lot of painful debugging. Curious about the learning curve for non-engineering founders who need to spec out systems but aren't deeply technical on the architecture side.

My first thought was that this seems extremely useful for games like satisfactory and factorio! Great work.

Interesting use case. How are people actually using the interactivity side, are they engaging with quizzes and navigation or mostly just watching passively?

Hi Product Hunt 👋

I’m Steve, founder of Libertify.

We built Libertify around a frustration we kept seeing everywhere:

important documents (reports, decks, PDFs) get sent… and then disappear into a black hole.

No engagement, no feedback, no real understanding.

So we asked: what if documents could actually work?

Libertify turns them into interactive experiences with AI, video, and real-time insights — so you know what’s read, understood, and acted on.

We’re just getting started, and your feedback means everything.

Would love to hear what you think 🙏

Most tools tell you what people skipped. This one tells you what they misread. That gap matters a lot when you're sending a financial plan or a proposal where one wrong assumption kills the deal.

How do you detect misunderstanding? Is it based on re-reads, time patterns, or something else?

Good luck today!

Here's Windsurf 2.0, the biggest Windsurf launch yet!

Here's the TLDR:

• Devin is now available in Windsurf, so you can delegate your work to cloud agents which can work even after your laptop is closed.

• Introducing the Agent Command Center, one place that lets you manage all your agents - local and cloud - from a single Kanban view.

• Spaces help you stay in the flow by grouping agent sessions, PRs, files, and context for a project. When you return to a Space, you can pick up where you left off.

the kanban view of agent sessions is clever — curious how it handles context isolation when you're bouncing between 10+ agents. does each session get its own scratchpad or is there a shared memory layer?

built-in Devin with its own VM is wild. delegating a task and literally closing your laptop while it works feels like science fiction. curious how well it handles codebases with complex dependencies - does it spin up the full dev environment automatically?

the kanban view for managing multiple agents is clever. we've been juggling Claude and cursor for different parts of our codebase and it gets messy fast. can you pause/resume individual agents or do they all run until completion?