PH热榜 | 2026-04-01

Hey Product Hunt 👋

This Easter, we'd love to invite you to try something a little fun.

If you're thinking about sending your friends a different kind of holiday surprise,

why not turn a simple greeting into a "voice egg"?

On Noiz, you can:

🥚 Open a voice egg and instantly unlock a randomly generated voice

🐰 Apply festive transformations to your voice and turn it into a bunny, chick, lamb, or other Easter-inspired characters

🎨 Design your own Easter voice by typing a prompt or uploading an image. Describe a fluffy bunny, a playful cartoon chick, or any character you imagine, and Noiz will generate a unique voice to match

You can use it for:

• Holiday greetings

• Voice messages to friends

• Social content

• Creative experiments with AI-generated voice styles

For example，

when you send:

"Happy Easter!"

It could sound like:

🐥 A cheerful little chick

🐰 A soft and cuddly bunny

🎭 Or a custom voice created from your imagination, just upload an image or describe it promptly.

Take part in our Egg Smash Event and unlock surprise rewards, from coupons and credits to exclusive voice styles and gift cards.✨💫

Sometimes, a tiny change in voice can make an ordinary message more fun and more memorable.

Come try your voice egg 👉https://noiz.ai/voice/design?ref=producthunt&utm_source=producthunt&utm_medium=easter_event&utm_campaign=voice_design_promo

And tell us in the comments ：

Which voice egg would you open first? 🐰🐥

Super cool launch! Do you guys see Noiz becoming more of a creative social tool or a creator platform?

Do you have any recommended social media templates for sharing these voice greetings?

Huge congrats on the launch, Vega @vega_chan & the Noiz team @Noiz AI and what a perfect time to crack this open! 🥚

I’ve been playing around with the “Egg Smash” mechanic, and honestly, it’s genius. It turns a simple voice note into a delightful little surprise, like a digital Easter egg hunt. The idea of uploading an image to generate a voice is super creative — I tried it with a photo of my grumpy cat, and the result was hilariously accurate! 😂

Here’s a small product thought from a fellow builder: I noticed the “random egg crack” is pure luck. What if you added a “Hint” feature? For example, after cracking, you could reveal a clue like “This voice has a British accent” or “This character loves carrots.” It would build a bit of suspense and make people want to “collect them all” to complete the set. It’d be a great gamification layer to keep users coming back after the holiday buzz.

Really excited to see where you take this! 🚀 Which egg would you say is the most unexpectedly popular so far? The bunny or the chick? 🐰🐥

Changing the emoji to switch between bunny, chick, and lamb voices worked seamlessly. The tone shifted instantly, and it sounded so natural.

The emoji metaphor for changing voices makes experimentation feel natural and fast. I lost track of time playing with it!

Love how simple this is .No setup no complexity just crack an egg and get a fun expressive voice instantly.

Voice eggs make festive expression feel less technical overall.

I've tested a lot of TTS tools, but Noiz AI's Easter update is by far the most creative and fun I've seen. The "voice egg" concept is genius, and the ability to transform your voice into Easter characters is a game-changer. It's not just about generating speech; it's about creating joy and connection.

What I find most impressive about Noiz AI's Easter edition is the seamless integration of fun and functionality. The "voice egg" is playful and engaging, while the custom voice creation is powerful and versatile. It's a rare combination that makes this tool stand out from the crowd.

It is refreshing to see holiday joy treated as a core product goal.

This Easter, Noiz AI has given us a new way to express ourselves and connect with the people we care about. Whether you're a creator looking to make unique social content or just someone who wants to send a fun greeting to a friend, Noiz has something for everyone. The "voice egg" feature alone is worth checking out!

What I love most about Noiz AI's Easter edition is how it turns a simple "Happy Easter" into a storytelling moment. You can make your greeting sound like a playful cartoon chick, a gentle lamb, or any character you can imagine. It's not just about sending a message; it's about creating a memorable experience for the people you care about!

Noiz AI's Easter update is a testament to the power of AI in making communication more expressive. The ability to generate a custom Easter voice from a prompt or image is incredible. You can describe a fluffy bunny, a playful cartoon chick, or any character you imagine, and Noiz will bring it to life with a unique voice. It's like having a creative partner for your holiday messages.

Easter voice eggs turn generic greetings into memorable surprises.

Noiz AI has truly thought of everything with this Easter launch. From the fun "voice egg" concept to the practical custom voice creation tools, every feature is designed to make your holiday greetings more personal and engaging. It's clear that they understand how important it is to connect with others in meaningful ways, even during the holidays.

This AI is a game-changer for holiday greetings! I’m so tired of sending the same generic “Happy Easter” texts to everyone. With this tool, I can make every message feel totally unique, tailored exactly to the person I’m sending it to. From cute, whimsical bunny-themed lines to sweet, sincere notes, every greeting feels like a special, handwritten gift. 10/10, would recommend!

It was simple to iterate on different emotions and character voices without touching any complicated settings.

If you're looking for a way to make your Easter greetings stand out this year, look no further than Noiz AI. The "voice egg" feature, Easter character transformations, and custom voice creation tools are all incredible, and the "Egg Smash Event" adds an extra layer of excitement. It's the perfect tool to make this Easter one to remember.

I tried the voice egg feature earlier, and it was such a fun surprise! Unlocking a random voice made my Easter greetings feel way more special.

Heheh! Love the vibe and my kids gonna love it even more. Congrats on the launch team!

I like how the output feels expressive rather than flat narration! Even a simple greeting has so much personality.

"Expressive voices in seconds" — what does expressive mean here specifically? Emotion range, character accents, speaking style? The difference between a voice that sounds "playful" and one that actually commits to a character is pretty significant.

This is a fun concept! How customizable are the generated AI voices - can users tweak tone, pitch, or style?

Will you be adding more prompt templates to help new users create better custom voices?

Hi everyone!

The engineering in Ollama v0.19 is a massive leap for anyone running local models on macOS. Moving to Apple's native MLX framework changes the game for performance, leveraging the unified memory architecture and the new GPU Neural Accelerators on the M5 chips.

v0.19 now also supports NVFP4, which brings local inference closer to production parity, and the KV cache has been reworked with cache reuse across conversations, intelligent checkpoints, and smarter eviction. For branching agent workflows like @Claude Code or @OpenClaw , that should mean lower memory use and faster responses.

If you have a Mac with 32GB+ of unified memory, you can pull the new Qwen3.5-35B-A3B NVFP4 model and test this right now. Running heavy agentic workflows locally just became a lot more viable!

Been running Ollama since like v0.12 and the speed improvements keep blowing my mind. The MLX integration is huge for M-series Macs tbh.

Smarter cache reuse is the underrated feature here. I run a coding assistant locally and switching between projects used to basically cold start every time. If the KV cache actually persists across sessions that changes everything for agent workflows.

The MLX rewrite is the real deal — been running Qwen3.5 locally on my M4 and the speed difference vs the old GGML backend is night and day. Cache reuse across conversations is clutch for agent loops too.

Finally, MLX-native inference. I've been running local models on my M2 Air for quick prototyping when I don't want to burn API credits, and the speed difference on Apple Silicon matters a lot when you're going back and forth between coding and testing. Curious how it handles the bigger models now, like 70B+ quantized. Does the memory management play nicer with other heavy processes running?

This is huge for local-first AI workflows. Curious how much real-world speedup people are seeing on M-series chips

Well done! Do all the current models work automatically with MLX with this version on macOS, or do you need to download a specific version of each model?

Will have to try this out as a previous version totally drowned my 16gb mini.

Hey Product Hunt! 👋
I'm Nikhil from Future AGI, and I'm excited to share traceAI with you today.

The Problem We're Solving
If you're building with LLMs, you know the pain: your agent made 34 API calls, burned through your token budget, and returned the wrong answer. You have no idea why.
Existing LLM tracing tools force you into a new vendor dashboard. But most teams already have observability infrastructure - Datadog, Grafana, Jaeger. Why add another?

OpenTelemetry is the industry standard for application observability, but it was designed before AI existed. It understands HTTP latency. It has no concept of prompts, tokens, or reasoning chains.

What traceAI Does???
traceAI is the proper GenAI semantic layer on top of OpenTelemetry. It captures everything that matters in your AI application:
- Full prompts and completions
- Token usage per call
- Model parameters and settings
- RAG retrieval steps and sources
- Agent decisions and tool executions
- Errors with full context
- Latency at every layer
And sends it to whatever observability backend you already use.

Two lines of code:
from traceai import trace_ai
trace_ai.init()

Your entire GenAI app is now traced automatically.
Works with everything:
- Languages: Python, TypeScript, Java, C# (with full parity)
- Frameworks: OpenAI, Anthropic, LangChain, LlamaIndex, CrewAI, DSPy, Bedrock, Vertex AI, MCP, Vercel AI SDK, and 35+ more
- Backends: Datadog, Grafana, Jaeger, or any OpenTelemetry-compatible tool
- Actually follows GenAI semantic conventions. Not approximately. Correctly. So your traces are readable in any OTel backend without custom dashboards or parsing.
- Zero lock-in. Your data goes where you want it. Switch backends anytime. We don't even collect your traces.
- Open source. Forever. MIT licensed. Community-owned.
We're not building a walled garden.

Who Should Use This???
AI engineers debugging complex LLM pipelines
Platform teams who refuse to adopt another vendor
Anyone already running OTel who wants AI traces alongside application telemetry
Teams building agentic systems who need production-grade observability

What's Next???
We're actively working on:
- Go language support
- Expanded framework coverage
Try It Now
⭐ GitHub: https://shorturl.at/gKG7E
📖 Docs: https://shorturl.at/AlyjC
💬 Discord: https://shorturl.at/v4llu

We'd love your feedback! What observability challenges are you facing with your AI applications?

Much needed! Since you’re positioning traceAI as a semantic layer over OpenTelemetry so do you see this becoming a standard like OTel itself or staying a developer-focused tool?

Open-source LLM tracing is exactly what was missing.

I run Claude API calls in a Celery worker — two calls per job,

one at temperature=0 (deterministic analysis),

one at temperature=0.7 (generative rewrites).

Right now I log both manually with structlog.

But correlating a specific trace across the two calls

when something fails in production is still painful.

Does traceAI handle multi-step pipelines where the same job

triggers two separate LLM calls with different parameters?

The OTel-native approach is the right call here. Most LLM tracing tools force you into a new dashboard and a new vendor relationship. The fact that this routes to Datadog, Grafana, Jaeger means teams can use what they already have instead of adding yet another pane of glass to monitor.

Curious about one thing: how does traceAI handle tracing across multi-agent workflows where one agent calls another? Do the traces compose into a single parent span, or do they stay isolated per agent?

Congrats on the launch.

Hey TraceAI team, great product. was able to get started by giving claude your documentation in a single day. We use this with our internal grafana server so it was a small setup but loving it! thanks!

The OTel native approach is the right call imo. Every time I've tried an LLM observability tool it wants me to install yet another dashboard and I'm already drowning in Grafana tabs lol.

Two lines of code to instrument is bold. Does it handle multi-step agent chains well? Like if I have a LangChain agent that calls tools that call other models, does the trace show the full tree or does it flatten everything?

Two lines is impressive but curious - how does it handle agent decision tracking when you have nested tool calls 3-4 levels deep? Running a bunch of AI agents for project management workflows and the traces get messy fast. The GenAI semantic conventions piece is what's interesting here - most OTel solutions just treat LLM calls as HTTP and you lose all the context about what the model was actually doing.

How does the Trace AI handles long running tasks or loops apart from standard loops? does it have any reasoning steps added to it?

Really enjoyed building this solution for AI pros. It gives you a clear look at how your AI agents are performing without any vendor lock in

This is the one of the best open source open telemetry solution out there, no vendor lock in and one stop solution. Great Work team!

this is going to be a one stop solution for anyone who is building agents and exploring agentic architectures!

GenAI observability has been broken for too long. TraceAI gets it right and this is the kind of observability layer every AI team needs but rarely has. Smart to make this open source and build trust first. Congrats team! 🚀

Since this is fully OpenTelemetry-native, I assume it should work seamlessly with backends like SigNoz as well?

If yes might try it there too seems a cool tool

This is awesome. I used to use a watch that would tell me when I needed to take breaks, but it wasn't for me. I currently use blue light blocking glasses, so this seems like another great health layer to add on.

Congrats on the launch! Any plans to bring Zzzappy to Windows or is it staying Mac-only for now?

Finally something that tracks actual input load instead of just a dumb timer. I code 8+ hours a day and my wrists only start hurting during heavy refactoring sessions, not when I'm reading docs. The frosted glass break screen is a nice touch too — way better than the ugly gray overlays other tools use.

As someone who builds a focus timer app, I respect this. Most break reminders are just dumb countdowns that interrupt you mid-flow. The fact that you're tracking actual arm input load and not just time is smart. I always ignore the basic "take a break" popups, but if it knows I've been hammering the keyboard for 2 hours straight that's harder to dismiss. Does it detect idle time or just raw input volume?

Its feel like scam, copied Lookaway and even didn't buy domain for website. After purchase lifetime deal you will vanish?

RSI prevention is underrated - wrist issues took out a few engineers on my team for weeks, costs more in lost output than any break schedule ever would. Curious how the timing works: is it adaptive based on input intensity, or fixed 20-20-20 style intervals?

This should be also for smartphones (which is even worse case, because people have smaller screen to focus on + they hunch over)

Oh man the arm input tracking is such a smart angle. Every break reminder I've tried just does a dumb timer, like yeah cool thanks for interrupting me mid flow for no reason lol.

The fatigue score based on actual keyboard/mouse usage is way more useful. Does it learn your patterns over time? Like if I tend to death grip my mouse during crunch weeks does it adjust thresholds?

Congrats on the launch! The arm input load monitoring is the part most tools miss entirely, screen time is easy to track but RSI builds up in ways you don't notice until it's already a problem.

Curious whether it adapts break frequency based on the type of work, heavy typing vs mostly reading, feels like it should trigger differently.

The features are great.

Hey Product Hunt! 👋

I'm Vivek, and I built ClawMetry because I got tired of not knowing what my AI agents were doing.

I run several OpenClaw agents. They handle code, research, deployment, scheduling. But every time one took 10 minutes on a task, I had no idea: is it stuck? Did it hallucinate? Is it burning through tokens?

NemoClaw (NVIDIA's AI agent sandbox) made running agents safer. But the built-in TUI is ephemeral and terminal-only. You can't see what happened yesterday. You can't watch 10 sandboxes from your phone. You can't track costs across your fleet.

So I built ClawMetry for NemoClaw. One command on the host, and every sandbox gets full observability:

🧠 Brain tab: every thought, tool call, and decision in real time
📊 Token tracking: per call, per session, no surprises
🔐 E2E encrypted: keys never leave your machine
🌐 Cloud dashboard: monitor everything from any browser

It's open source (MIT), free for local use, and took about two months of obsessive building.

With your love and support, ClawMetry has been downloaded 95,000+ times across 100+ countries. This NemoClaw integration is the next step.

What's coming next:

• Policy drift detection (get alerted when sandbox policies change)
• Remote egress approvals from your phone
• Fleet-wide policy management

Cloud sync is $5/sandbox/month. Local dashboard is free forever.

Would love your feedback. Happy to answer any questions!

🔗 https://clawmetry.com/nemoclaw

Congarts to the launch.

@vivek_chand The E2E encryption approach is smart since most monitoring tools force you to ship sensitive agent data to their servers. Quick question though, how does the real-time visualization perform when you're monitoring multiple sandboxes simultaneously?

The real-time flow visualization sounds like exactly what I needed when debugging agent chains that would mysteriously stall for minutes at a time. :D

Congrats on shipping @vivek_chand! I'm curious about the policy drift detection feature you mentioned, how would that work in practice when sandboxes update their own policies?

Quick question about the token tracking per session feature, does it break down costs by specific tool calls or just aggregate session totals?

This reminds me of debugging microservices but for AI agents. Makes me curious, how the real-time flow visualization handles really chatty agents that make tons of tool calls..

Awesome builder, awesome product :D

perfect timing - the Codex team just reset the usage limit for all plans! source

THIS IS AMAZING

Self-hosting the relay -realistically, how involved is that setup? Asking because "self-host if you want full control" sounds great until it's 11 pm and you're fighting a Docker config:)

Started using Remodex in the first couple of days. I chose to pay the subscription instead of going for the opensource version. Purely on how clean & smooth the app is.

Now I can ship while I sh*t 😂

can you also build for android

👋
Hey PH,

Since day one, we've been building Clarify as the autonomous CRM that actually does work for you — not just stores it. Today, we shipped an entirely new way to use Clarify. One that helps you not just detect, manage, and close deals, but find more of them.

Here's the problem we kept hearing: sales teams spend more time managing their outbound stack than actually doing outbound. Find leads here. Enrich there. Sequence over there. Then, hope the CRM syncs correctly. It's a full-time job for something that should be automated.

With Clarify, you can now prospect with a prompt and outbound in a click. All in your (autonomous) CRM.

Here's how it works:


🔍 Lead Finder
Describe the leads you want. Clarify builds a targeted, enriched list with company data and decision-makers.


✉️ Campaigns
Describe what you want to send them. Clarify builds single or multi-step sequences individually tailored to each lead on your enriched list. Make direct edits or ask Rep, your personal sales agent, to make them for you. You’re in control.


⚡ Pipeline managed
New deals are automatically detected from replies. Tailored follow-ups get written while you sit back and relax as your pipeline grows.

All this happens in one streamlined motion. In one tool. It sounds too good to be true, but it isn't.

See for yourself and try it for free → clarify.ai/propecting

Honestly so proud of the team today as we launch Prospecting on PH. 💚

Prospecting has been months in the making, and what made it different was how deeply we involved real customers throughout. We ran structured testing sessions with founders and GTM teams, watched how they used it, listened hard, and iterated fast. The product you're seeing today is genuinely shaped by that feedback.

One thing I'm especially excited about: we leaned heavily into Rep, our AI assistant, to power the core experience. Describe the leads you want, describe what you want to say - Rep handles the rest. It sounds simple, but seeing it click for users in testing was something else. We really believe this is where CRM is headed, and Prospecting is our strongest signal of that yet.

A personal highlight - I used Campaigns to run most of our outreach during the testing phase. Wrote a prompt, got a sequence, sent it. The quality of leads we were able to find and the speed at which we could reach them genuinely surprised me. It made the whole process feel effortless in a way I hadn't expected.

If you're a founder or part of a GTM team spending too much time juggling tools, give it a try. Find your ICP from 100M+ contacts, run campaigns, and watch deals get created automatically as prospects engage - all without leaving your CRM.

Would love to hear what you think!

As a stakeholder in this, I've been close to this build as it came together.

I can attest that Lead Finder + Campaigns makes outbound simpler and more effective.

From a GTM standpoint, it’s a faster path from “here’s the ICP I want” to actually getting in front of them with something personalized.

I’ve been using it to find and reach the exact founders I want for a few upcoming dinners. Tight targeting, relevant outreach, and less overhead.

If you’re running outbound, worth trying.

Happy to share how I’m using it or compare notes.

I rarely comment on PH launches. But this one deserves it.

Two months ago, I started using Clarify, and it blew me away. Clarify's UX is unparalleled. It's intuitive and has everything I need. From the Chrome add-on to the built-in workflows.

The new Lead Finder helps me eliminate one tool from my toolkit, meaning everything lives in one place.

Well done, team!

Prospecting tools needed to be rebuilt from the ground up and I’m excited to see this new launch from a team who delivers the best CRM. Congrats team!

the "no more sourcing from one tool, enriching in another, sequencing in a third" problem is real and most teams are duct-taping these workflows together with zaps and frustration. collapsing it into a single CRM that actually understands context is well done.

the hard part will be lead data quality. autonomous outbound that runs on autopilot works beautifully when targeting is right, and breaks badly when it's sending "personalized" emails to people who've moved companies, changed roles, or aren't actually buyers. the tools for finding leads are ahead of the tools for qualifying them, and that gap tends to show up in reply rates pretty fast.

Clarify offers so much out the box against incumbants like pipedrive and Salesforce. The AI features are very promising. Most attractive is how light weight it has been.

Congrats on the launch. As a founder my self, even with automation, I use many tools for my cold email outreach channel. NGL most lead filter don't really work, I always QC the list with Claude. This is truly an incredible concept. Excited to check it out!

👋 Product Hunt

The bet we made building Clarify: founders shouldn't need three tools to run their pipeline. Find the right people, reach out, close them... all in one place.

Seeing it actually work for customers makes the last few months worth it.

Happy to answer anything. ask away.

I’ve been using Clarify for over 8 months now and have already killed my Hubspot subscription. Looking forward to seeing how I can now cancel the mess of lead gen tools! Looks promising so far. Can’t wait to see the results.

Veo 3.1 Lite is Google’s most cost-effective video generation model, now available via the Gemini API.

It solves the biggest bottleneck in AI video — high costs — by offering <50% pricing compared to Veo 3.1 Fast, without compromising on speed.

What makes it different is this strong cost-to-performance balance, enabling high-volume video creation.

It supports Text-to-Video and Image-to-Video, 720p/1080p resolution, flexible aspect ratios (16:9, 9:16), and adjustable durations (4s, 6s, 8s).

Perfect for developers building scalable video apps where efficiency and iteration matter.

I have made hours of videos on my YT channel with previous versions, I tried today 3.1 and all I can say it's, amazing job, still.

Hello Product Hunt!

Oh boy, was this one fun to make.

I wanted to bring the magic of generative AI to an audience that love expressing themselves in stickers and cute visuals.

Usually, you'd need to mess around with prompts to get the exact output you want.

Or be left with presets that don't look the part.

Snapstick makes creating beautiful stickers easy:

1. Pick or take a photo

2. Select from 6 predefined styles

3. Customize your sticker, and share it, or use it to cause chaos in your group chat.

Snapstick also lets the user create Sticker Packs they can share with friends.

Members of a Sticker Pack can also add stickers to the pack, which syncs across all members.

Remix lets you create different variations of your stickers.

Add text, expressions, poses or props to create reactions fit for every occasion.

Make It Real brings your sticker into the real world.

Print it as a kiss-cut sticker, get it on a mug, or a tote bag.

You can even order your sticker on a t-shirt.

The perfect gift, or personal treat.

Snapstick is available around the world.

Make It Real is available in Europe and the US.

Thank you for reading, and I hope you'll join me in making the world a bit more delightful.

We sure need it.

—Mads

congrats on the launch mate!

Hey Product Hunt, I'm Tahir, co-founder and CTO of OpenBox AI. Today we're thrilled to introduce OpenBox, the trust platform for agentic AI that makes enterprise grade governance available to everyone.

AI agents are now operating across workflows, systems, and organizations at scale. The question every team building with agents faces is the same:

• How do you know what your agents are doing

• How do you prove they acted within policy

• How do you meet compliance requirements without rebuilding your entire stack
  

OpenBox answers that. It delivers runtime governance, cryptographic verification, and enterprise grade compliance at the point of execution, enforcing identity, authorization, policy, and risk across every agent action and cross system interaction.

OpenBox integrates via a single SDK with no architectural changes to your existing stack. It works natively with LangChain, LangGraph, Temporal, n8n, Mastra, and more.

You get:

• Production grade SDK

• Cryptographic audit trails

• OPA based policy engine

• Built in runtime guardrails

• Dynamic risk scoring

• Human in the loop controls

• Full observability from day one
  

We built OpenBox on the belief that trust should be a right, not a privilege. Every organization deploying AI agents deserves the same governance and accountability infrastructure, whether they are a startup or a regulated enterprise.

That is why the core platform is available in production, with no usage limits and no credit card required.

Would love to hear from everyone building with AI agents today:

• What are you building

• How are you handling governance

• What is missing in your stack

Happy to answer everything here 👇

Very good to see you guys live, How are you handling policy enforcement across different agent frameworks without adding latency?

Congrats on the launch BTW 🎉

Super excited to see OpenBox live. Would really appreciate any thoughts and feedback.

The scale at which AI agents are being deployed today makes this the right moment for OpenBox. Runtime governance, cryptographic verification, enterprise-grade compliance - available to every organisation, from day one. Proud to be part of this.

I've been dealing with audit nightmares from our ML ops team and that OPA policy engine integration could actually save us months of compliance work it seems.

What Tahir has laid out here is what we have been building toward: a platform that governs every agent action at the point of execution, with full observability and cryptographic proof, from day one. If you are building with agents and want to understand how it works technically, happy to answer everything here.

Stoked to see this launch, the cryptographic audit trails piece is what really caught my attention here. How do you handle the performance overhead when you're signing every single agent action in a high-throughput environment?

Huge congratulations @natsuda_uppapong @phaituly @tonyopenbox on shipping this. How does the cryptographic verification works when you need to halt an action mid-execution, does the signature still get created for the attempted action that got blocked?

I'm wondering how the cryptographic verification works when an agent pulls from multiple data sources with different permission levels in a single workflow?

Cryptographic verification of agent actions is the interesting piece here. What exactly is being signed — the prompt, the tool call, the output, all of the above? And when you say 'verify,' is that post-hoc audit trail or can you actually halt an action mid-execution if it fails a policy check?

Do you think openbox or other similar tools in future will become a standard layer in every agent stack, like auth or logging today?

Nice work on this. How does it integrate with existing agent frameworks like LangChain or similar tools?

@tahir_mahmood8 Congratulations. And happy product launch.

I've been thinking about this space a lot lately and honestly most governance solutions I've seen are either too heavyweight for dev teams or just basic logging that doesn't actually prevent anything bad from happening.

How does this handle the performance hit when you're doing real-time policy checks on every agent action, especially for high-frequency workflows where latency actually matters?

@tahir_mahmood8 @asim_ahmad_cfa @grover___dev Congrats on the launch... lets presume you were to explain this product to someone with minimal technical knowledge as it relates to use case within a business (a business that uses AI but isn't too deep into the compliance / governance side of how this works) - how would you go about outlining the use case.... asking for a friend!

This is a problem that does not get enough attention yet. Everyone is focused on making agents more capable, but the question of "how do you prove they acted within policy" is going to matter a lot more as agents start touching real workflows at scale.

The cryptographic verification angle is interesting. Most governance approaches I have seen are audit logs after the fact. Proving compliance at the point of execution is a different thing entirely.

Question: how does OpenBox handle governance for agents that are pulling context from multiple systems with different access policies? For example, an agent that reads from both a public knowledge base and a restricted HR system in the same workflow. Does the governance layer enforce per-source permissions, or is it more at the action level?

Love the direction here. Are you targeting enterprise use cases first or keeping it flexible for smaller teams as well?

Hi everyone!

You have a new teammate in @Slack.

Slackbot can now work across your conversations, files, apps, and agents in one place. Ask it for context, have it prep a meeting, route a task, or handle the next step, all without the usual tab switching.

Maybe it is time to see whether your new AI teammate can pass its probation review?

@zaczuo Do you plan to make this AI bot available for the Business version? Because as far as I can see, you currently need the Pro+ version, and that’s quite a significant cost when you consider the whole organisation and the number of paid tools we've already used.

---
@patryk_iwaszkiewicz almost like our internal bot Iwan :D

Meeting prep and report analysis in Slack flow makes sense. The hard part is always context - can Slackbot actually read channel history to understand what the team has been discussing, or is it limited to the current thread? That determines whether it helps with real async coordination vs just being a smarter search box.

Having an AI agent that actually lives inside Slack where all the context already is makes way more sense than switching to some separate AI tool and copy pasting stuff over.

The meeting prep angle is interesting. Does it pull from Slack threads and channels to build context or does it need separate integrations with calendar apps?

Meta just launched prescription-optimized Ray-Ban Meta AI glasses: Blayzer Optics (Gen 2) and Scriber Optics (Gen 2), starting at $499, perfect for the billions needing vision correction without sacrificing smart features.

Everyday glasses wearers want all-day AI smarts (like hands-free nutrition tracking via voice/photo logging with personalized recs, US-only) but hate bulky add-ons, these are lightweight, optician-adjustable frames built for comfort and nearly all prescriptions.

What's new? First truly optical-forward Ray-Ban Meta designs with overextension hinges, interchangeable nose pads, and tailored fit, plus fresh colors/lenses for Ray-Ban Meta (e.g., Shiny Transparent Grey with Transitions Sapphire) and Oakley Meta (e.g., Prizm Rose Gold).

Key features & benefits:

• Nutrition tracking for healthier choices;

• WhatsApp summaries/recall (EAP, on-device privacy);

• Neural Handwriting rolling out everywhere (even iMessage, write on any surface);

• pedestrian nav expanding US-wide; more like display recording and widgets coming soon.

• Keeps you intuitive, helpful, phone-free.

For daily AI users, nutrition trackers, WhatsApp power-users, walkers; ideal for logging meals, quick message catch-ups, discreet replies, turn-by-turn guidance.

The prescription lens support is the obvious unlock here — most people who actually need glasses all day have been blocked from smart eyewear because of this. But how does the prescription integration work practically? Do you go through a licensed optician, upload your Rx somewhere, or is Meta partnering with specific lens providers?

Is the problem with data leakage fixed?

Congrats on your launch! Being able to export transcripts is super useful. Can you detail how transcription works? Several questions:

Does it use YouTube’s native subtitles, or generate its own transcript?

What languages are supported for transcription and AI processing?

How does it handle videos with multiple speakers or heavy background noise?

Can I edit the transcript manually before exporting it as TXT/PDF?

Some times the videos are so long and we just need the context and move on, this kind of tool is really helpful when the presenter takes ages to get to the point!

Congrats on the launch @debojyoti452 love the product. Early adopter and have manged to use it as my go to email as a service solution for FoundersBoxx and other projects. Love the new features.

positioning email infrastructure as a product team problem rather than a backend one is the interesting move here. delivery visibility and sandbox testing are usually buried behind DevOps access, so making them something the product team can reach directly removes the whole "can someone check if that onboarding email actually fired" loop, which anyone who's shipped a transactional flow knows is more annoying than it sounds :)

This hits home.

I’ve used SendGrid in the past, and honestly, the biggest friction wasn’t just setup, it was everything after.

Deliverability felt like a black box, and when things went wrong, support wasn’t always helpful or timely. You end up spending more time debugging email than actually building your product.

What stands out about Keplars from this story is the focus on:
- reducing setup friction
- making email feel like part of the product (not separate infra)
- and actually giving visibility into what happens after sending

That last part is huge, because sending is not aka delivery

Congrats for the launch!

Hey Product Hunt! I'm Charlie, founder of Stiinks.co. We've been heads down for the past year building what we believe is the worst link-in-bio platform on the internet. Every Stiinks page ships with clashing color palettes, broken links, intrusive ads, endless popups and analytics that somehow go negative. We're incredibly proud of what we've built and we'd love your feedback on how to make the experience even more unusable!

PS. Happy April Fools from Liinks.co, the link-in-bio that actually works

There’s something interesting about showing the anti-product to highlight the real one.
Feels like a more honest way to differentiate in a crowded space.
Do users actually spend time exploring Stiinks, or is it more of a quick laugh and bounce?

I haven't seen so many colours in one place in my whole life. :D

Hi,

I built this because running multiple Claude Code agents across multiple IDE and terminal windows was getting messy, it was all changing quite fast, and nothing out there handled it the way I wanted. I needed one place to see all my agents, review their changes, and spin up new ones without constantly switching between windows.
I've been building Baton from within Baton for a while now, which has been a pretty fun loop. Would love to hear what you think!

Awesome man! What’s the long term vision for Baton like do you see it becoming an “OS for AI developers” or more like a power tool within existing workflows?

This looks really close to what I've been wanting in an agent-driven development tool. Does it (or could it in the future) support orchestrating agent sessions on remote devices? Similar to using VSCode on Windows to connect to remote Linux environments via SSH.

the git-isolated workspace per agent is the detail that makes parallel agents actually practical. without it you're just tab-managing concurrent terminals that will eventually collide on the same files, and the mental overhead of tracking what each agent touched defeats the point. the isolation isn't an implementation detail, it's the whole model.

I wouldn't recommend installing this app but it seems worth a conversation... it's a privacy and security nightmare, as demonstrated by two developers who reverse-engineered both iOS and Android apps and found serious problems:

• Excessive data collection. Both platforms bundle Firebase Analytics, track device identifiers, and send your IP and device info to Google’s servers. The Android version goes further — requesting 22 permissions including camera, microphone, exact location, contacts, call logs, calendar, and the ability to read and send SMS. None of this is necessary for a government news app.

• Remote control via feature flags. The app checks a remote endpoint for “feature flags,” meaning its behavior can be changed server-side at any time without an app update. What it does today may not be what it does tomorrow.

• Poor security practices. API keys (Firebase, Google, ad unit IDs) are hardcoded into the binary. The Android version uses React Native with outdated dependencies carrying known vulnerabilities. One reviewer called the codebase “held together with mass-produced duct tape.”

• Bottom line: This app collects far more data than it needs, can change behavior silently, and is built to a lower standard than most side projects. Just use the website.

Don't be an April Fool. Avoid this app!

Fellow Mac dev here — notch integration is such a clever use of that space. Clean UX.

Using the notch as actual UI real estate instead of just hiding it is a nice move. I always thought that space was wasted. Open source too? Might peek at the repo to see how you're handling the notch interaction, that's a fun macOS API challenge.

the token cost tracking per project in the menu bar is more useful as a signal than a budget tool. a session that costs 10x the average is almost always one where something went wrong, not one where you accomplished 10x more.

the secret detection scanning for leaked API keys is the feature that quietly saves someone's day. session histories can end up in sync directories or version control if you're not careful, and most people don't think about what's in them until after.

Hey Product Hunt, I'm Liran. I built Claudoscope because I manage Claude Code rollout across engineering teams at my company and had no way to tell what sessions were costing us or what was happening inside them.

The existing menu bar trackers scrape your session cookie from claude.ai. That doesn't work for Enterprise API deployments. Claudoscope reads the local JSONL session files directly from ~/.claude/projects/, so it works on any setup, Enterprise included.

What actually got me started: I found my database password sitting in plaintext in a session file. Claude Code had read a .env and echoed it back as a tool result. Nobody was checking for that. So now the app scans for leaked credentials using entropy-based filtering and alerts you when something shows up.

Beyond secret scanning, it gives you:

Session history browser with a chat-style viewer and full-text search

Cost analytics per project and per session (supports both Anthropic and Vertex AI pricing)

Config health linter with 19 rules that checks your CLAUDE.md, skills, and hooks (found broken skill metadata and oversized configs across our team that nobody had noticed)

MCP server and skills browser

It's native Swift/SwiftUI, not Electron. Runs locally, no telemetry.

brew tap cordwainersmith/claudoscope && brew install --cask claudoscope

Free and MIT-licensed. If you're running Claude Code on an Enterprise plan where cost visibil

The ability to draw directly on screen is super useful. Can you detail the drawing tools?