Check out this blog post about our latest AppControl release that works with your preferred AI agent, adds Chinese/Turkish languages, and much more! Give it a try and let us know how you like the update.
Just download AppControl from our home page, and install it on top of the previous one. No uninstall is necessary.
This is amazing! I installed it using option A for Claude and stumbled around not really understanding what I was doing or what to expect.
I did not have Claude Desktop installed and had never used Claude, so that was the first thing. But it went fine.
Then continued with my stumbling, but just followed the instructions. It took me a bit to find “Settings” in the Claude desktop app (had to click on the profile icon in Claude), but once I found that and just blindly continued following the instructions, everything proceeded.
I wondered, well, what now? And found the “Example prompts” in the instructions and tried a few.
Thanks!
Thanks so much! Yes, it was surprising to me too when I first used it. You can ask pretty crazy stuff like “why was my fan ramping up while my PC was locked yesterday as I walked up to it” or “what are the countries my most used binaries are from” and then now we added idle time, so you can ask about things your PC was doing while it was idle.
One thing to remember is that the longer AppControl is on your PC, the more useful the data is. Because if you just installed AppControl, then start using this… it doesn’t have much history to check so it isn’t as useful. You need a couple days for AppControl to be installed to be more useful because then it has the information in its stats on your hard drive (local) to read off of.
I am guessing that different AI tools may produce different answers?
I went with the option to use Claude simply because it seemed to be highlighted more and perhaps easier to install.
Is there any advantage to using one of the other “MCP-compatible AI tools and agents”?
I presume (certainly hope!) I don’t have to understand anything in the “Available Tools” section of the instructions?
If one were to decide they don’t want to use this feature, is it recommended to remove the extension from Claude? Or is this never going to do anything if AI integration is turned off in AppControl?
Yes, different AI tools can provide different answers, however… they both read the same data so the answers on the data should be similar unless the AI is hallucinating or having some kind of problem.
As far as an advantage of using the other AI tools, it’s user preference. There are some things you could do with an AI Agent tool that has the ability to read our data. For example, maybe the agent could be instructed to occasionally look if certain processes are active or not, and make a report, or maybe the agent could check how the server or PC is doing so it knows the resource usage is reasonable. I personally don’t use an agent on a PC currently though so we aren’t even fully aware of everything that’s possible, but we were hoping maybe some PC users with agents might try it?
No, once Claude desktop is set up you should be good. That’s what I am using myself.
Yes, if you don’t want this feature you can remove the extension from Claude, but also please note that Claude asks your consent to access the data, so you could just not give consent going forward and it can’t see the AppControl data anymore. And since it asks consent, it should not look at your data anymore.
I hope the helps? Thanks for being one of the first people to try it out!
By the way… the AppControl MCP server also allows integration with self hosted fully offline LLM models/agents for privacy conscious people through for example ollama (https://ollama.com/) + webui (https://openwebui.com/) and while we haven’t tried it ourselves yet, it should give an option to query everything fully locally without using any cloud at all if someone prefers.
Using Claude for Windows Desktop, I quite quickly ran into a dreaded “You are out of free messages until 3:00 PM”. This was after asking only a couple of questions.
It would cost me CA$300/year to upgrade to Claude Pro and I am a cheap SOB. For now, I am guessing I will see how many queries I can do for free every day (or whatever).
I really don’t understand all this stuff and am stabbing in the dark.
I have the Perplexity.AI desktop app installed and have never had it tell me I had to wait a period of time because I wasn’t paying for it, so I asked it, Is there an “Module Context Protocol” ability with Perplexity.ai?
It turns out there is one, but currently only for the Mac and currently only for subscribers, although they say it will be coming to the free tier in the future.
Here’s hoping for an AI tool that will allow a lot of free queries every day. Otherwise, this may be a fascinating and powerful feature I am just not willing to pay for 
Yeah, Claude does cost money and I agree that’s not awesome. I wonder if tomorrow you could use Claude to help you set up something like https://ollama.com/ that is free? It may be an option for you if you have pretty powerful hardware because it would just run locally on your PC, and maybe that would just be free?
Or, look at the OpenAI software option (and others) we have on Github and see if those do more for free?
The nice thing is that MCP is an open standard and our server is open source, so you should be able to find an AI that will operate it, hopefully in an affordable way. This is all super new, so that’s probably why there is no Perplexity Windows version yet, but I bet it’s coming soon!
We give Claude as the first example though because it’s pretty easy and fast to try, and it already has a large user base, but you are absolutely not limited to only Claude.
@CTaylor we are considering adding a guide on how to use Ollama and other free alternatives to Claude, etc… that way people don’t have to pay to use this. We’ll try it ourselves first though and see how well it works before recommending that.
I think that would be great to have guides to use other AIs. the instructions for using Claude were straightforward and easy to follow, but looking at Ollama…without some pretty explicit instructions, I am afraid of just ending up with a non-functioning mess! 
Yes, I will admit I did not try Ollama myself. One thing to think about though is how you can use another AI to set that up and figure it out that way.
For our own MCP server page, we plan to release a revision of the instructions shortly that will explain how to set it up in a much easier way.
@kaser sorry for the issue. Is this due to an update, or is this the first time you are installing AppControl?
Do you use any third party security tools/software? Of course, Windows defender/security is fine. Maybe something is not letting us install?
It appeared during the update, then I deleted the program and tried installing it for the first time, but the same problem
only Windows defender
@kaser do you have AppControl installed on multiple Windows user accounts on your PC? Or, do you only have one Windows account with AppControl installed?
@AppControlOfficial only one Windows account
Darn… we thought that could give us a clue to what was going on. OK, I’ll ask our team and see if I can get some more questions to ask so we can get to the bottom of this issue.
Works ok with LMstudio too, with the standalone .exe and the provided json. When I say works, it actually depends on the model. Gemma 4 fell into a loop constantly querying the database and never giving a result. Nemotron worked fine though. So your results will vary depending on the model.
One question though. Why is it constantly trying to phone home? It’s been doing this since the initial release and quite incessantly (every few minutes). I have “check for updates” off and I don’t appreciate apps that have hidden telemetry that is not mentioned anywhere, much less provide a way to turn it off.
@randombit why is what trying to phone home? Do you mean AppControl itself?
It’s because we have “suspicious alerts” (optional which you can turn off) and “insights” that tell you what your executables are. We have to send the insight information or hashes of your executables to our server to check them and if you are looking at data on executables then it’s going to need to contact our server every time. If we tried to put all of that on our installer it would be pretty massive and also out of date pretty fast.
We provide an optional privacy mode you can use instead if you prefer Privacy Policy - AppControl but you won’t have insights or suspicious app detection.
Or are you talking about our MCP extension itself? The source is on our Github and it shouldn’t do that?
Thanks for trying out the MCP with LMstudio, that’s good to know! We’ll see if we can do something to improve Gemma 4 or if it’s something outside our control.
Thanks!
I mean the app itself. I just have it running in the background, don’t use any insights or anything and it keeps trying to contact various servers constantly. You can understand why I find this suspicious. I hadn’t really noticed the insights feature since it’s blocked by the firewall and doesn’t work anyway.
Thanks for letting me know about the privacy feature. I reinstalled it and will get back to you if it still tries to connect. Regarding the insights, maybe you can have a link that opens the page in the browser in case someone wants to checkout a file. Instead of just saying the feature is not available.
