Perplexity just made OpenClaw without the security vulnerabilities

I want to love OpenClaw… I really do. But I just can't bring myself to accept the security vulnerabilities and complicated setup that comes with it. Perplexity Personal Computer? Now that's something I can get behind, as it might finally deliver my dream of a proper AI personal assistant.

OpenClaw finally delivered AI's long-awaited promise of a real-life Jarvis

Security protocols optional

OpenClaw took the world by storm a month or so ago, quickly becoming one of, if not the most starred project on GitHub in history. A few weeks later, the developer of OpenClaw joined the team of OpenAI . But, the OpenClaw saga isn't all butterflies and rainbows.

OpenClaw might be an awesome piece of software, but it's plagued by two big problems: it's complex to set up, and its security is questionable at best.

I want to love OpenClaw . I've tried multiple times to use it, but I just can't bring myself to keep it turned on for very long. The setup is simple at first, but complicated to get going. It's expensive to use with decent models, and profoundly useless with more affordable models. The security issues are where I really have a problem, though.

Because the project started as a fun vibe coding program, the developer didn't focus on security at first. By the time OpenClaw was massive, it was almost too late to fix it. Still, over a month from it blowing up in popularity, there are many security vulnerabilities that have yet to be patched. The platform is just too insecure for me to trust with my day-to-day workflow, and many are starting to believe the same as I do.

I'll admit, when OpenAI hired the OpenClaw developer, I really was hoping that a ChatGPT version of OpenClaw would be rolled out soon thereafter. Sadly, it hasn't happened yet, and there's no indication of if or when. Thankfully, other companies are stepping up to fill the void.

Perplexity's Personal Computer could be just what we've been waiting for

Full audit trails, a kill switch, and more make Personal Computer a compelling personal AI agent

A few weeks after OpenClaw captured headlines, Perplexity decided to launch a similar agent called Perplexity Computer . Designed for a multi-agent workflow in the cloud, Perplexity Computer offered a lot of the functionality everyone wanted from OpenClaw, but without the security vulnerabilities that OpenClaw came with.

However, what Perplexity Computer lacked was local system access. You see, one of the biggest selling points of OpenClaw was the fact that it ran on your hardware, so it could access your files, apps, and programs natively. This is a huge feature, and one that many people used heavily—Perplexity Computer couldn't do that.

Just this week, Perplexity launched Perplexity Personal Computer . Unique naming scheme, I know, but it's still an awesome update to Perplexity Computer.

With Perplexity Personal Computer, you have all the benefits of Perplexity Computer—like a multi-agent and multi-model workflow, orchestrations, deep research, and so much more—but running on your own system.

While Perplexity Personal Computer can't use local LLMs and still relies on cloud-based models, it does have a lot of unique and very security-focused features in it. For starters, there's a full audit trail of everything it does, so if something goes haywire, you can go back and see just where it went wrong.

My favorite function of Perplexity Personal Computer though is the kill switch—a simple way to just stop everything that the AI is doing in case it starts to really go off the rails.

AI is a fantastic tool , but it's definitely not without its drawbacks. Just a few months ago, Google's Antigravity agentic IDE decided to delete a guy's entire D:\\ drive on his computer with no way to recover the data. So, if Perplexity Personal Computer, which runs on your local machine with access to your local files, starts to do things it shouldn't, you can easily just make it stop— something that OpenClaw wasn't good at respecting .

I'm just waiting for the perfect personal AI assistant that actually does what I need

Is it really too much to ask?

We're in a weird middle area of AI. It's smart enough to be useful, but the tools haven't quite caught up yet. I am really waiting for an AI that can function like a true personal assistant, but without complicated setup, expensive tokens, or security vulnerabilities.

Really, I just want to be able to have a summary of my day gathered by an assistant, looking at my calendar, various task and project management apps, email, and Slack, and have that summary delivered to me at a specific time each day.

Then, I want to be able to ask pointed questions about what I should do with my day at that time and get relevant answers. You know, standard secretary and assistant stuff, but without hiring a personal secretary or assistant. That shouldn't be too much to ask.

---

Personal AI assistants aren't far off, but they're still too far away for my liking

I know that OpenClaw technically delivers on this promise, but I've already aired my grievances with it and why I won't rely on it. I really hope that Perplexity Personal Computer, or another AI platform, can deliver on this promise sooner rather than later. Until then, I'll keep fumbling around with tools that promise to solve my problems, but really just waste my time.