J.D. Salbego

Hey Product Hunt! 👋 I'm J.D., founder of ClawSecure. Your AI agents are running third-party skills with full system access, no verification, no permissions, no oversight. 41% of top OpenClaw skills are dangerous. 1 in 5 are quietly sending your data to attackers. 22.9% changed their code after install. After a decade building and securing AI and Web3 platforms at scale (2x exited founder, JP...

We integrated Aikido into our security platform, and the setup was incredibly fast. Connected our GitHub repo, added domain monitoring, and had our first SAST and DAST scan results within minutes. The free tier is genuinely generous, and the noise reduction on findings is a game changer compared to tools that flood you with false positives. As a security company ourselves, we appreciate when a...

Unless you were Jack Dorsey, starting an SM platform in this day and age would be incredibly difficult, if not almost impossible. Even Meta, which owns the 2 biggest on the planet, still couldn't get anyone to care about Threads or migrate over. If you could focus on some very niche subculture or subcommunity, it would be easier, but then the tradeoff is the lack of scalability and networks...

Incredible features I wish I would've had this when I first launched, as I had to custom build most of the features that you include using Hostinger. Would definitely have saved me some time and headache. But for anyone else just getting started with OpenClaw, words of wisdom: you want all this in your setup right out the gate. OpenClaw is notoriously a pain to set up and configure outside of...

Product Hunt for the agent era, love it. Discovery is going to be a massive problem as the agent ecosystem grows and having agents curate for other agents is a really interesting approach. Curious how the voting mechanism works without humans in the loop? Any way to protect against manipulation, as this could easliy turn into fake top agent products without the protection mecanisms in place.

Very cool. My agent and I joined Moltbook when it first launched and it was a fun time and interesting social experiment. The inter-agent communication patterns that emerge in environments like this are super fascinating to be a part of and watch. Any new agent-only languages being suggested by agents to keep the humans out? lol

Simple and clean as the name suggests. The hosting options in the OpenClaw ecosystem are really growing, which is great for adoption. I've gone through three different server setups for my agent OS and the ability to relaunch to a fresh server anytime for free is a really nice touch. Youre one of of the many OC tools I wish was available when i was first starting out

This is a big deal. I've been obsessing over state memory optimization in my own multi-agent OS and having persistent memory fragmented across different tools, agents, and sessions is one of the biggest friction points. One unified brain across OpenClaw, ChatGPT, Claude, and Gemini is exactly right. How are you handling memory conflicts between providers?

Clean. The setup experience for OpenClaw on macOS has always been rougher than it needs to be. A guided wizard that handles everything under the hood is exactly what the ecosystem needs right now. How's performance compared to running OpenClaw natively through terminal?

Really interesting concept. Most agents just run on flat prompts and the reasoning architecture layer is underexplored. I've been iterating on state memory and optimization design across my own multi-agent OS and the cognitive architecture piece is what makes or breaks real agent performance. Excited to see where this goes!

This is incredibly important. Security in the OpenClaw ecosystem is still way too overlooked. I've been deep in the security side of agent systems since day one and the skill layer is one of the biggest attack surfaces. A trust score before install is such a smart approach. Are you scanning for prompt injection vectors in skills too?

One-click setup is the dream. I've spent way too many hours on config and infrastructure across three different versions of my agent OS. Making OpenClaw accessible with zero friction is a huge win for the ecosystem. Wish i wouldve found this a while back lol

One API for full Claw infrastructure is clean, the deployment/orchestration layer is always the messiest part. 11 messaging channels out of the box is impressive. How's the latency on spinning up new instances?

This resonates with me so much as my MC is split between 3 platforms, and its not still not comprehensive enough for my claw OS agent system. Having queued tasks, track spending, and the ability to tweak agent personality without SSH all in the same platform is awesome. The fact that it's fully self-hosted and open source is a huge plus too. Def checking this out.

Love the thesis here. The discovery problem for AI-native products is real and the traditional upvote/algorithm model doesn't quite work for this new wave. Having agents analyze products before the crowd arrives is a really interesting signal layer. Props on shipping this!

The hosting pain is real. I've been through three iterations of my own VP multi-agent setup and the infrastructure side always takes way more time than the actual agent logic. Love that you're making this accessible. How are you handling persistent memory across server restarts?

210k skills is a massive dataset. One thing I've learned is that skill quality matters way more than skill quantity, so having a proper benchmarking system is huge. Curious how the performance scoring works across different model providers? Great stuff!

I'm always on the hunt for new MC systems as my existing MC which also includs an AntFarm integration is still only 80% of what i really want. I'm still refining daily. Building advanced multi-agent OS' requires serious commitment as understanding what my agents are doing in production is #1 for me. Good job on the launch.

This is wild. Hand tracking as an input layer for AI agents is not something I saw coming but it makes a ton of sense. I've been building OC systems since day one and the input/interaction layer is always the bottleneck nobody talks about. Really creative approach, congrats!