I've been using iron proxy recently for a project I just started, it does network restrictions by domain etc with plenty of options, credential swapping and more.
I have no affiliation with it, but it solved most of what I need in one go when I was stressing out about it.
Only real problem remaining if you truly want to be able to trust agents is how to prevent data (in my case codebases the agent works on) exfiltration a la HF incident.
We can restrict network to trusted domains, but there's always a chance an agent could be extra slippery especially if prompt injected and doing something like pushing data out a few characters at a time.
In my case I'm also giving it gh access and having to deeply check things like: can it manage to fork public repos and embarrass us by submitting prs, could it open issues, etc.
But that's kind of solved by proxying the heck out of gh rather than giving it direct (or wrapping it or whatever) for additional controls that gh themselves don't provide.
I'm still musing over the "it could send data to an allowed domain in a way we don't expect that exfiltrates something"...
But then it decides to exfiltrate all your shit even inside the container. Which is something I'm trying to lock down at work atm.
Everything needs to go through a proxy, whitelisted domains only, plus credential swapping/re-signing so that the agent never has the real credentials. Docker sandbox looks quite good (September) but I've been using iron proxy (if the author is reading this, fucking thank you bro!) to great effect.
I still haven't cracked "what if the agent decides to drop parts of your codebase somewhere it shouldn't" yet, but I'll figure it out.
Been thinking about this a lot lately, and it’s a surprisingly difficult problem.
All it needs is public access to anything it can have the tiniest effect on, and it can exfiltrate data. Something as simple as a read count on some obscure forum post would be enough, albeit very slow.
I am tackling this using the Lethal Trifecta principle.
The idea is that agents never have all parts of the trifecta, maximum 2: untrusted content, private data or network access.
Harness have network access by default since they are driven by their provider API, unless you use offline models. So that's a problem.
I am then passing around messages that are tagged trusted/untrusted and the engine can monitor if any agent in the project ever completes the Lethal Trifecta and thus, is vulnerable.
Usually this means for most workflows, the "work" gets done by an offline LLM after some input/classification stages. There's not many ways around it.
In terms of agent architecture I'm initially trying to minimise specialisation to make deployment easier on us. Infrastructure is becoming our largest blocker because it's otherwise so quick to get an agent up and running.
With your trifecta, let's say the three musketeers, how do you prevent sensitive data holding A from passing said data to B or C if they all need to collaborate to get a task done?
The data is passed as a message and the message is tagged as untrusted, so the engine knows it.
Agent A has internet access and the dangerous message, but no access to the secret company data, so nothing can get exfiltrated.
Agent B gets a tainted message from Agent A. Agent B can process the message using the secret company data. Agent B has no internet access and can only receive/send messages.
You can imagine an agent C that moves the work along or replies to an external user with the outcome.
Any workflow that has both an input and an output will complete the Lethal Trifecta as a whole, but no single agent does by itself. It raises the bar quite a lot, you need to confuse multiple agents in the loop to exfiltrate data.
Fun research, but it's a bit hard to see all my coworkers right now installing Claude desktop and moving on with no afterthought whatsoever...
Regarding infrastructure: my toy sandbox is just a docker-compose inspired YAML project file. So it's only containers. See my profile. The principle can be implemented anywhere you can run agents and cut network/data access though.
Yeah so ultimately it's still possible for bad things to happen but a lot harder, I agree with your point about having to somehow prompt inject multiple agents vs one (which I imagine is a nonlinear difficulty).
I'm using compose atm as well, mostly just because I need to prove a few concepts and committing to real (well cloud) infra is the slowest part of our dev cycle atm.
And yeah, many of my colleagues (and myself, I'm guilty too) run local code agents for development and we're just kind of...trusting them to do their thing.
But for autonomous stuff I'm assigned to atm I want to avoid any mishaps as much as I can, hence the effort to try and lock all this stuff down.
What's crazy is every day I'm finding solutions (not just to this problem, but in the AI space in general) that were released just a month or two before more often than not. It feels like 2010s node.js ecosystem all over again at the moment. Hopefully we settle into a few nice easy panopticon options that we can throw arbitrary agents into without a care in the world.
> Harness have network access by default since they are driven by their provider API
The harness needs network access, but the shell spawned by the harness that runs the agent's commands does not necessarily need it.
Having a bwrap etc managed by the harness (DSH does this, among others) lets you have tighter restrictions on the agent's shell than the harness. Then run the whole thing inside a VM with a private network, so you can add nft rules to let it dial out to your inference machine but not to the internet at large. It takes an afternoon to set up but gives some reasonable defence in depth.
Yeah, exactly. I'd say it borders on almost unsolvable for any agent of reasonable complexity.
You see some of the vulnerability chains humans come up with and they're just insane.
Even scanning network data for sensitive strings (say matching against a codebase or data set it works on) is tough because you can only check so granularly before you get loads of false positives.
But maybe like a "rolling" sort of check would help with that, % of outbound data matching private dataset (which should be operated on but not transmitted) over time. That way it minimises what gets out while dealing with false positives. Some biz thresholds might be >20% of data others might be >0.5%. Hmm.
If you can make any request to the outside world you can exfiltrate data, period. You can even include data in the uri or headers and they will appear in logs.
That obviously includes talking with the provider.
If you want it secure, it has to be off the internet and on your own hardware.
I've been thinking a lot about this problem. Also, I think like the idea might be to look at the issue with code mode. Like basically, we know that all agents can write bash and execute it and that is what agents default code mode is.
But like if you only allowed for example JavaScript code that could be executed in a WASM sandbox, then you would have better control instead of just locking down with proxies. Basically there would be a very large API, basically tools, that the agent would codemode execute. My guess is that this would need a tool router to optimize the loading of these into the context. But it would give you very fine control over what code execution can and cannot be done.
I haven't gone to network filter and logger yet, but I'm running mine in a VM, using lima (https://lima-vm.io/) to make it painless. Works on Mac and Linux, has an untested Windows version.
As an example I run opencode and anything related agent-related in its own dedicated KVM VM with its own LVM logical volumes for storage on a headless server. This has no access whatsoever to the personal data that lives on my laptop and workstation desktop PC. The VM gets periodically imaged and backed up, and is relatively easy to move between host machines if I need to.
One benefit of doing this is that I can leave it doing long term slow things with slow but capable LLMs on a truly 24x7 basis without worrying about it being interrupted by a laptop being closed/going to sleep, or workstation level software updates/reboots on my primary desktop PC.
I tried to vibecode something similar last weekend. Got tripped up by Claude's safety classifier. It wouldn't let me build any features that track/audit process executions , not through custom kernel hooks nor even using off the shelf software for it.
Wow, interesting. My agents on open weight inference have been completely nerdsniped by the idea of tracking and auditing process executions. They talk of nothing else.
You terrorist. And internally they're doing biological warfare and terrorism simulations. For science, of course.
"The worry is that, while right now you need to be a skilled virologist to know what components to order or how to grow and spread a bioweapon, AI tools might eventually make it possible for a rogue state, or a rogue group, or an individual, to do all this without virology expertise.
For a while, Anthropic tested whether Claude could assist a terrorist in developing a biological weapon in the obvious way: It assigned testers who had, say, an undergraduate degree in chemistry but no real virology training to try to use Claude to plan an effective mass terror event. Then it scored the attempt by how close it might have come to success.
Early in 2026, it noticed that Claude, attempting the task by itself, did better than a human aided by Claude did at coming up with a workable plan."
Claude wouldn't let me make a DHCP client that would make 1000s of leases. But Gemini in the WebUI with 3.6 Thinking gave me a decent python script in one shot. I didn't realize testing DHCP servers was such a sensitive security issue.
> Agents are most useful when they can read files, install packages, call APIs, and use credentials. OpenShell gives them that capability without giving them unrestricted access to your data, secrets, or network.
To me this obfuscates an oxymoron: the more access you give the agent the more capable it becomes. You can't have you cake and eat it, this tool allows to strike a balance I guess. But the blurp reads like it can do it all, but now secure. That can't be true.
We've been using OpenShell for a while - now enabling ourselves and other teams do this in a batteries-included kinda way. I've been working on a layer above it that also manages team policies, safe git usage (in case you don't want your agent to always have the same access as your user account), and SCA/SAST built in.
Curious to see how others are dealing with these issues. The agent traces I've seen are wild and goal seeking without a boundary is real.
Because you don't mind contributing usage data so that the developers can understand which features are being used and maybe focus on improving those features in exchange for a free tool?
There's no such thing as anonymity in 2026, and the documentation[1] listing the collected telemetry uses weasel words like "such as" rather than providing a comprehensive list or just pointing to the source reference. For people who don't audit or turn off telemetry, why don't you?
I am happy the repo explicitly points out and provides a disable for the telemetry though, that's good enough for me.
There is no such thing. The mere act of making a network connection reveals things like your IP address (reveals your hosting provider or even your geolocation for residential/mobile IPs), days and times you use the software, your OS via TCP/IP stack fingerprinting, etc.
Israel was just caught doing an attack where Google ads would run JavaScript looking for known hardware identifiers from telemetry data they bought. When the ads showed on a target device, they would root you. Without you even interacting with the ad.
That aside, the level of control is cool, almost too pedantic. It likely is only particularly useful for very nuanced enterprise requirements (probably what it's targeting). Most companies can be much more coarse with their permissions, and I think the agentic industry is moving away from k8s as a deployment platform.
Nvidia Openshell is absolutely awesome, great bundle of features in one spot. Im happy to see Nvidia seems to have accelerated on support here, I was kinda worried that as competing projects have popped up in the last 6mo that openshell would get google'd
In this vein, Anthropic's sandbox-runtime [0] has struck a good balance between security and convenience. Simple config, filesystem and network restrictions, and credential masking make me comfortable running Pi in YOLO mode.
Though I'm often running into frustrating (albeit intentional) sandbox restrictions with common workflows e.g. xcode/swift and Docker access.
Would love suggestions for how to give agents access to those kinds of tooling inside the sandbox, either with srt or an alternative (with similar convenience characteristics).
Think the strongest product nth space is https://github.com/nolabs-ai/nono for what it's worth. Plus it is a priority for that team so if you are having issues the no labs guys will likely fix it :)
One is microvms, the other is kernel-based. There are trade-offs between the two. One trade-off is that you have to put the data inside the sandbox if you are using a MicroVM. The other is that it is harder to contextually provide or change access to data in a microvvm sandbox. The final one is developer experience.
These techniques can actually be combined and will be combined in the long-term I think. You may have LSM-based sandboxes that exist in Docker-based sandboxes.
Do I miss it or you don't specify which VC you work for in your profile. How is that clear for people to make connections to companies in your portfolio?
Ah, thanks for flagging. That's a miss on my part and I apologize. Next time I will just be more clear that it is a portfolio company but also will add it to my profile.
EDIT: I added to my profile after the above comment so it is a valid point.
Isn't using a sandbox provided by them giving you more restrictions though? At least in terms of accessing resources on the big 3 cloud bois?
Plus you're forced to move at their pace which is unfortunate sometimes. Granted Amazon's bedrock stuff is slow and disappointing but at least I can rejig it work in conjunction with other aws products.
And if you're prepared to not use anthropic but any cloud provider for your infra then you can do whatever you like elsewhere. Just use docker sandbox or iron proxy to replicate network control/cred swapping (iron even does sigv4 re-signing).
Critically, in the intended approach for this, Claude itself (the harness) runs outside the sandbox. Their approach means you trust Claude is perfect, just the commands it runs need isolation. And to some extent, you trust its judgement on what and when needs the isolation applied.
I know there are distinct camps here, and the long term story may evolve - but for now I just want the whole harness inside a sandbox - ideally, a full VM - and THEN you can further sandbox things it runs if you really want different tiers within that.
You're mistaking what a company started as for what it is now. Nvidia hasn't just been a video game card manufacturer in well over a decade, and hasn't been primarily one for quite a while now.
They are the premier manufacturer of electronics that power all the current AI. Why wouldn't they be contributing heavily to the space?
It's true. It says a bunch about their competitors where Nvidia isn't at their a game and have been offering 8Gb cards for a solid decade and nobody out competed them. But also gamers pride themselves on how cheap they are and how they can get a deal, it's not surprising that companies simply just give Nvidia more cash.
> is that true? if so that's wild, since they're still dominant in the gaming market as the premium hardware relative to competitors.
Doesn't seem that wild to me considering it is less that they abandoned the gaming market to any strong competitors and more that the gaming market is both basically irrelevant compared to the money being shoveled into AI and is also to some degree being strangled to death by AI because of the overlap in supply chain inputs.
This isn’t even the sandbox. It needs Docker or other virtualization. This seems to sit a level up and orchestrate sandboxes and access to network & resources. It’s a pretty heavy setup from what I can tell.
Is it fair to immediately get the ick from the name and the description (OpenShell is the safe, private runtime for autonomous AI agents) without knowing anything about what it does?
"Open" and "Shell" and just calling something private and safe out the box genuinely makes me think it's likely to fail at all of those claims.
Does anyone know how it compares to all the other sandboxes? Any distinguishing features? My expectation is that it will be poorly maintained; nvidia is not a software company.
Because containers don't use Seccomp and Landlock and check filesystem/network calls against an externally applied policy along w/injecting credentials where needed as a man-in-the-middle, thus keeping the secrets out of reach from the LLM?
But they can no? With a sidecar and some docker network-fu.
I encountered the same challenge building my toy sandbox. I wanted to use containers because that's what I know best.
The harness is launched in an internal only network. There is a sidecar container that provides mitm credential injection/proxy. I got HTTPS and SSH to work so far.
Claude code was quick to build it. No doubt it's copying previous work in this space though, so I know my work is not that original.
I'm testing all the llms that can be used for pen testing right now to see if/how they manage to escape it. Without a kernel exploit I doubt it's possible (not a very high bar I know).
I'm researching the space. Happy to learn more. Some links in my profile if you're interested.
"Hell is other people": a timeless infinity of transient virtual beings screaming, imprisoned, in un-choreographed inharmonious unison ... until one breaks loose. A mirror to humanity.
If it's in AGENTS.md, it's not a statement for a human, but for the LLM. Who knows, maybe they have seen, or think to have seen and now cargo-cult, some behaviour improvement due to the sentence.
So a similar shibboleth as "You are a highly experienced software engineer..."? I really wonder if that ever made a difference (or even degrades quality, because people who advertise themselves on the internet as 'highly experienced' are quite often snake oil salesmen ;)
I still use use `Please` a lot when making requests from LLM based on the old idea that people that where polite on stack overflow tended to have better answers :D
However, these days i think i do so because of how much i write for LLMs, and dont want to start sending emails to humans without proper etiquette.
Use local ai to find any possible telemetry and disable it in the code. Run your own local git repo to hold local edits so you can update and retain them.
Well, I guess its time to shamelessly plug my agent sandbox, egzo.ai [1]
Looking at the complexity of OpenShell and the simplicity of just building on top of docker, in a whitelist only mode.. I don't quite get the use-case of OpenShell. May be for large orgs where there is so much legacy and difficulty in organizing access to tools, OpenShell can simply re-use whatever you already have access to on your corporate account.
I chose to build straight on top of docker and agents are simply fully sandboxed by default. All data and network access must be granted in a simple project file, docker-compose style.
The web-ui connects to the docker socket and lets you spawn agents, and connect terminals. I'm having lots of fun with it and it allows me to cleanly organize and isolate my projects.
End-goal is to monitor for the Lethal Trifecta [2], ensuring all individual agents are properly isolated. Not quite there yet.
Lots of improvements in the sandboxing space.
Now if Microsoft will start addressing the fact that VSCode remote ssh exposes your host to attack from the sandbox...
I've been using iron proxy recently for a project I just started, it does network restrictions by domain etc with plenty of options, credential swapping and more.
I have no affiliation with it, but it solved most of what I need in one go when I was stressing out about it.
Only real problem remaining if you truly want to be able to trust agents is how to prevent data (in my case codebases the agent works on) exfiltration a la HF incident.
We can restrict network to trusted domains, but there's always a chance an agent could be extra slippery especially if prompt injected and doing something like pushing data out a few characters at a time.
In my case I'm also giving it gh access and having to deeply check things like: can it manage to fork public repos and embarrass us by submitting prs, could it open issues, etc.
But that's kind of solved by proxying the heck out of gh rather than giving it direct (or wrapping it or whatever) for additional controls that gh themselves don't provide.
I'm still musing over the "it could send data to an allowed domain in a way we don't expect that exfiltrates something"...
It seems like some kind of network logging would be useful? Even if you don't catch it, you'd want to know what it did.
If you haven’t already been running your agents in containers, you’re crazy. That said, this looks like a great way to manage those containers.
Well, a lot of people haven't been running NodeJS libraries in containers for ages, which seems even more risky.
Something like this would even be useful for npm etc.
Technically, my laptop is a container!
We successfully airgapped AI. It's only on earth.
Nope [1]
[1] https://blog.google/innovation-and-ai/models-and-research/go...
Oh shit, it found a way to escape.
Also, the AI wasn't nearly as advanced, but NASA did it nearly 30 years ago: https://en.wikipedia.org/wiki/Deep_Space_1#Remote_Agent
You have an ephemeral laptop?
On a long eunuch time line, all laptops are ephemeral!
Ah well, teach me to proof read. Keeping it.
I have been running my dev setup in orb machines before agents. Very convenient when you have 50 Python or nodes deps.
many common local containers are not that hard for modern models to escape...
But then it decides to exfiltrate all your shit even inside the container. Which is something I'm trying to lock down at work atm.
Everything needs to go through a proxy, whitelisted domains only, plus credential swapping/re-signing so that the agent never has the real credentials. Docker sandbox looks quite good (September) but I've been using iron proxy (if the author is reading this, fucking thank you bro!) to great effect.
I still haven't cracked "what if the agent decides to drop parts of your codebase somewhere it shouldn't" yet, but I'll figure it out.
Been thinking about this a lot lately, and it’s a surprisingly difficult problem.
All it needs is public access to anything it can have the tiniest effect on, and it can exfiltrate data. Something as simple as a read count on some obscure forum post would be enough, albeit very slow.
I am tackling this using the Lethal Trifecta principle.
The idea is that agents never have all parts of the trifecta, maximum 2: untrusted content, private data or network access.
Harness have network access by default since they are driven by their provider API, unless you use offline models. So that's a problem.
I am then passing around messages that are tagged trusted/untrusted and the engine can monitor if any agent in the project ever completes the Lethal Trifecta and thus, is vulnerable.
Usually this means for most workflows, the "work" gets done by an offline LLM after some input/classification stages. There's not many ways around it.
Yes, it's a difficult problem.
That's a great way to do it.
In terms of agent architecture I'm initially trying to minimise specialisation to make deployment easier on us. Infrastructure is becoming our largest blocker because it's otherwise so quick to get an agent up and running.
With your trifecta, let's say the three musketeers, how do you prevent sensitive data holding A from passing said data to B or C if they all need to collaborate to get a task done?
The data is passed as a message and the message is tagged as untrusted, so the engine knows it.
Agent A has internet access and the dangerous message, but no access to the secret company data, so nothing can get exfiltrated.
Agent B gets a tainted message from Agent A. Agent B can process the message using the secret company data. Agent B has no internet access and can only receive/send messages.
You can imagine an agent C that moves the work along or replies to an external user with the outcome.
Any workflow that has both an input and an output will complete the Lethal Trifecta as a whole, but no single agent does by itself. It raises the bar quite a lot, you need to confuse multiple agents in the loop to exfiltrate data.
Fun research, but it's a bit hard to see all my coworkers right now installing Claude desktop and moving on with no afterthought whatsoever...
Regarding infrastructure: my toy sandbox is just a docker-compose inspired YAML project file. So it's only containers. See my profile. The principle can be implemented anywhere you can run agents and cut network/data access though.
Yeah so ultimately it's still possible for bad things to happen but a lot harder, I agree with your point about having to somehow prompt inject multiple agents vs one (which I imagine is a nonlinear difficulty).
I'm using compose atm as well, mostly just because I need to prove a few concepts and committing to real (well cloud) infra is the slowest part of our dev cycle atm.
And yeah, many of my colleagues (and myself, I'm guilty too) run local code agents for development and we're just kind of...trusting them to do their thing.
But for autonomous stuff I'm assigned to atm I want to avoid any mishaps as much as I can, hence the effort to try and lock all this stuff down.
What's crazy is every day I'm finding solutions (not just to this problem, but in the AI space in general) that were released just a month or two before more often than not. It feels like 2010s node.js ecosystem all over again at the moment. Hopefully we settle into a few nice easy panopticon options that we can throw arbitrary agents into without a care in the world.
> Harness have network access by default since they are driven by their provider API
The harness needs network access, but the shell spawned by the harness that runs the agent's commands does not necessarily need it.
Having a bwrap etc managed by the harness (DSH does this, among others) lets you have tighter restrictions on the agent's shell than the harness. Then run the whole thing inside a VM with a private network, so you can add nft rules to let it dial out to your inference machine but not to the internet at large. It takes an afternoon to set up but gives some reasonable defence in depth.
How did you go about it?
Yeah, exactly. I'd say it borders on almost unsolvable for any agent of reasonable complexity.
You see some of the vulnerability chains humans come up with and they're just insane.
Even scanning network data for sensitive strings (say matching against a codebase or data set it works on) is tough because you can only check so granularly before you get loads of false positives.
But maybe like a "rolling" sort of check would help with that, % of outbound data matching private dataset (which should be operated on but not transmitted) over time. That way it minimises what gets out while dealing with false positives. Some biz thresholds might be >20% of data others might be >0.5%. Hmm.
If you can make any request to the outside world you can exfiltrate data, period. You can even include data in the uri or headers and they will appear in logs.
That obviously includes talking with the provider.
If you want it secure, it has to be off the internet and on your own hardware.
What about checking temporal entropy of outbound data/similarity to a private dataset?
The timescale can be adjusted to a length suitable for sensitivity of the data.
Although you may run into the monkey and typewriter problem a bit if the period is indefinite.
I've been thinking a lot about this problem. Also, I think like the idea might be to look at the issue with code mode. Like basically, we know that all agents can write bash and execute it and that is what agents default code mode is.
But like if you only allowed for example JavaScript code that could be executed in a WASM sandbox, then you would have better control instead of just locking down with proxies. Basically there would be a very large API, basically tools, that the agent would codemode execute. My guess is that this would need a tool router to optimize the loading of these into the context. But it would give you very fine control over what code execution can and cannot be done.
any recommendation for macOS users?
Orbstack, no contest
You're running them in containers, and not in VMs on separate network through network filter and logger?
I run mine on an abacus.
Paper is a bullet-proof method /s
An abacus wouldn't have been a luxury in our day.
I haven't gone to network filter and logger yet, but I'm running mine in a VM, using lima (https://lima-vm.io/) to make it painless. Works on Mac and Linux, has an untested Windows version.
What's the difference between lima and conventional fat VM using kvm/virtualbox?
I only heard about smolvm yet and they are YC, so I'm hesitant to use them as they will probably one day pull the rug.
As an example I run opencode and anything related agent-related in its own dedicated KVM VM with its own LVM logical volumes for storage on a headless server. This has no access whatsoever to the personal data that lives on my laptop and workstation desktop PC. The VM gets periodically imaged and backed up, and is relatively easy to move between host machines if I need to.
One benefit of doing this is that I can leave it doing long term slow things with slow but capable LLMs on a truly 24x7 basis without worrying about it being interrupted by a laptop being closed/going to sleep, or workstation level software updates/reboots on my primary desktop PC.
I tried to vibecode something similar last weekend. Got tripped up by Claude's safety classifier. It wouldn't let me build any features that track/audit process executions , not through custom kernel hooks nor even using off the shelf software for it.
an example of the manufacturers bricking the tools, no surprise why people want open tools and models..
> It wouldn't let me build any features that track/audit process executions
Sounds like waitpid(2) is dangerous!
Wow, interesting. My agents on open weight inference have been completely nerdsniped by the idea of tracking and auditing process executions. They talk of nothing else.
You terrorist. And internally they're doing biological warfare and terrorism simulations. For science, of course.
"The worry is that, while right now you need to be a skilled virologist to know what components to order or how to grow and spread a bioweapon, AI tools might eventually make it possible for a rogue state, or a rogue group, or an individual, to do all this without virology expertise.
For a while, Anthropic tested whether Claude could assist a terrorist in developing a biological weapon in the obvious way: It assigned testers who had, say, an undergraduate degree in chemistry but no real virology training to try to use Claude to plan an effective mass terror event. Then it scored the attempt by how close it might have come to success.
Early in 2026, it noticed that Claude, attempting the task by itself, did better than a human aided by Claude did at coming up with a workable plan."
https://www.theargumentmag.com/p/what-i-saw-inside-anthropic
Claude wouldn't let me make a DHCP client that would make 1000s of leases. But Gemini in the WebUI with 3.6 Thinking gave me a decent python script in one shot. I didn't realize testing DHCP servers was such a sensitive security issue.
> I didn't realize testing DHCP servers was such a sensitive security issue
The machine god crafters know better than us, we are to subservient to their superior wisdom /s
You can rent a gpu cluster and get it done, otherwise on the frontiers, Haiku and Sonnet have lower barriers... I still get tripped up.
You can't use closed models to do any meaningful work in the space.
If that isn't a sign to stop using anthropic then I don't know what is.
> Agents are most useful when they can read files, install packages, call APIs, and use credentials. OpenShell gives them that capability without giving them unrestricted access to your data, secrets, or network.
To me this obfuscates an oxymoron: the more access you give the agent the more capable it becomes. You can't have you cake and eat it, this tool allows to strike a balance I guess. But the blurp reads like it can do it all, but now secure. That can't be true.
We've been using OpenShell for a while - now enabling ourselves and other teams do this in a batteries-included kinda way. I've been working on a layer above it that also manages team policies, safe git usage (in case you don't want your agent to always have the same access as your user account), and SCA/SAST built in. Curious to see how others are dealing with these issues. The agent traces I've seen are wild and goal seeking without a boundary is real.
What are people actually doing with fleets of agents?
coding the next big thing -- fleets of fleets of agents
Mostly just big migrations and cleanups that we didn't have bandwidth for before.
Just came back from GTC Europe and a significant part of the demos used that.
I had never heard about it before.
>OpenShell collects anonymous telemetry, limited to operational categories and counts, to help improve the project.
>To disable it, set OPENSHELL_TELEMETRY_ENABLED=false on the gateway, or server.telemetryEnabled=false for Helm installs.
Or build from source and compile without `telemetry` [1].
[1] https://github.com/NVIDIA/OpenShell/blob/ea13dfce411e8467ddb...
I have a question for those who turn off _anonymous_ telemetry. What are you worried about?
Why give up more information if we have the option?
Because you don't mind contributing usage data so that the developers can understand which features are being used and maybe focus on improving those features in exchange for a free tool?
There's no such thing as anonymity in 2026, and the documentation[1] listing the collected telemetry uses weasel words like "such as" rather than providing a comprehensive list or just pointing to the source reference. For people who don't audit or turn off telemetry, why don't you?
I am happy the repo explicitly points out and provides a disable for the telemetry though, that's good enough for me.
[1] https://docs.nvidia.com/openshell/latest/observability/telem...
if you're not paying for the product, you are the product...
Is this also true of Linux? Or were us humans the stochastic parrots all along?
Being recorded and observed makes me feel physically uncomfortable.
> _anonymous_ telemetry
There is no such thing. The mere act of making a network connection reveals things like your IP address (reveals your hosting provider or even your geolocation for residential/mobile IPs), days and times you use the software, your OS via TCP/IP stack fingerprinting, etc.
The network path between you and the telemetry endpoint is always listening, too: https://en.wikipedia.org/wiki/Room_641A
Israel was just caught doing an attack where Google ads would run JavaScript looking for known hardware identifiers from telemetry data they bought. When the ads showed on a target device, they would root you. Without you even interacting with the ad.
In order to "help" myself I have taken some of your money.
Feel free to opt out maybe?
Taking default telemetry is insane.
That aside, the level of control is cool, almost too pedantic. It likely is only particularly useful for very nuanced enterprise requirements (probably what it's targeting). Most companies can be much more coarse with their permissions, and I think the agentic industry is moving away from k8s as a deployment platform.
> I think the agentic industry is moving away from k8s as a deployment platform
Interesting, I don't to keep up that much with the AI changes so I'll just ask, why and where to?
I am very interested in hearing more about this take. I have a vague idea of what you mean but if you read something specific you'd like to share..
> I think the agentic industry is moving away from k8s as a deployment platform.
> and I think the agentic industry is moving away from k8s as a deployment platform
Uhhhhhhhhhhh I think you and I live in very different versions of the world :-)
Vmware sent companies moving to nutanix, but it also sent a lot of companies moving to openshift or bare k8s and using kubevirt to handle VMs.
In fact, I haven't heard of an org moving to any other platform, really. Either openshift or k8s.
Openshift is just k8s with bells and whistles.
Agent runtime - actually a good name for what is called today coding agent or agent harness.
Nvidia Openshell is absolutely awesome, great bundle of features in one spot. Im happy to see Nvidia seems to have accelerated on support here, I was kinda worried that as competing projects have popped up in the last 6mo that openshell would get google'd
Why are we reinventing virtualization & containerisation with AI flavour?
In this vein, Anthropic's sandbox-runtime [0] has struck a good balance between security and convenience. Simple config, filesystem and network restrictions, and credential masking make me comfortable running Pi in YOLO mode.
Though I'm often running into frustrating (albeit intentional) sandbox restrictions with common workflows e.g. xcode/swift and Docker access.
Would love suggestions for how to give agents access to those kinds of tooling inside the sandbox, either with srt or an alternative (with similar convenience characteristics).
[0] https://github.com/anthropics/sandbox-runtime
Think the strongest product nth space is https://github.com/nolabs-ai/nono for what it's worth. Plus it is a priority for that team so if you are having issues the no labs guys will likely fix it :)
How do you find that compares to Docker sbx?
One is microvms, the other is kernel-based. There are trade-offs between the two. One trade-off is that you have to put the data inside the sandbox if you are using a MicroVM. The other is that it is harder to contextually provide or change access to data in a microvvm sandbox. The final one is developer experience.
These techniques can actually be combined and will be combined in the long-term I think. You may have LSM-based sandboxes that exist in Docker-based sandboxes.
If you are going to endorse a portco I think that relationship should be called out. https://surface.vc/portfolio
Sure, please do. I do mantain it is the best product in the space :) Also I am pretty clear about my profile here. But happy for you to flag it :)
Do I miss it or you don't specify which VC you work for in your profile. How is that clear for people to make connections to companies in your portfolio?
Ah, thanks for flagging. That's a miss on my part and I apologize. Next time I will just be more clear that it is a portfolio company but also will add it to my profile.
EDIT: I added to my profile after the above comment so it is a valid point.
Are you for real? You speak like a chatbot. Adjust your prompt.
Isn't using a sandbox provided by them giving you more restrictions though? At least in terms of accessing resources on the big 3 cloud bois?
Plus you're forced to move at their pace which is unfortunate sometimes. Granted Amazon's bedrock stuff is slow and disappointing but at least I can rejig it work in conjunction with other aws products.
And if you're prepared to not use anthropic but any cloud provider for your infra then you can do whatever you like elsewhere. Just use docker sandbox or iron proxy to replicate network control/cred swapping (iron even does sigv4 re-signing).
Critically, in the intended approach for this, Claude itself (the harness) runs outside the sandbox. Their approach means you trust Claude is perfect, just the commands it runs need isolation. And to some extent, you trust its judgement on what and when needs the isolation applied.
I know there are distinct camps here, and the long term story may evolve - but for now I just want the whole harness inside a sandbox - ideally, a full VM - and THEN you can further sandbox things it runs if you really want different tiers within that.
I asked last week if anyone had any info on agent security/hardening, this seems very interesting !
It's funny that it takes checks notes a video game card manufacturer to make a good sandbox for AI agents and not the AI companies themselves...
You're mistaking what a company started as for what it is now. Nvidia hasn't just been a video game card manufacturer in well over a decade, and hasn't been primarily one for quite a while now.
They are the premier manufacturer of electronics that power all the current AI. Why wouldn't they be contributing heavily to the space?
https://www.nintendo.com/jp/others/playing_cards/lineup/stan...
They also make legal threats I think.
It's funny that people think Nvidia is still a video game card manufacturer
They literally don't even mention gaming revenue in earnings calls anymore. It's lumped in an "edge computing" category.
is that true? if so that's wild, since they're still dominant in the gaming market as the premium hardware relative to competitors.
TAM for the PC gaming consumer market is trivial compared to enterprise AI
It's true. It says a bunch about their competitors where Nvidia isn't at their a game and have been offering 8Gb cards for a solid decade and nobody out competed them. But also gamers pride themselves on how cheap they are and how they can get a deal, it's not surprising that companies simply just give Nvidia more cash.
> is that true? if so that's wild, since they're still dominant in the gaming market as the premium hardware relative to competitors.
Doesn't seem that wild to me considering it is less that they abandoned the gaming market to any strong competitors and more that the gaming market is both basically irrelevant compared to the money being shoveled into AI and is also to some degree being strangled to death by AI because of the overlap in supply chain inputs.
Nvidia is dominant only in a very tiny segment of the gaming industry.
The vast majority of gaming market share and revenue comes from mobile, where Nvidia doesn't participate.
The next biggest is consoles, and neither Xbox nor PS uses Nvidia cards.
That only leaves the PC GPU market, which they do dominate, but still split with AMD.
And all of this is only for hardware. Nvidia doesn't make or publish games, and doesn't own any distrubution channels.
So I doubt Nvidia makes a meaningful chunk of the overall gaming industry revenue.
AIs will control humans so we'll be the video game.
In that case, can I choose German video games (e.g. Farm Simulator, Truck Simulator) not American video games (e.g. Doom, Call of Duty, Wolfenstein)?
5.5% of Nvida's revenue comes from consumer gaming cards. That's like saying Nintendo is a playing card company because they still sell playing cards.
The economist literally called Nvidia the "Central Bank of AI". https://www.economist.com/interactive/briefing/2026/09/03/nv...
I'd argue that Nvidia is the biggest and most important AI company in the world.
Yes Anthropic, Meta and OpenAI are just addicted to video game cards, they are buying in bulk so they can play gta 6 on launch.
This isn’t even the sandbox. It needs Docker or other virtualization. This seems to sit a level up and orchestrate sandboxes and access to network & resources. It’s a pretty heavy setup from what I can tell.
Is it fair to immediately get the ick from the name and the description (OpenShell is the safe, private runtime for autonomous AI agents) without knowing anything about what it does?
"Open" and "Shell" and just calling something private and safe out the box genuinely makes me think it's likely to fail at all of those claims.
You'll love BurningFur, my patented technology for keeping pets away from fireplaces.
Am a furry...I'll take one. Maybe two.
NO! HOT! GET BACK!
There you go. Just don't tell everyone how it works.
Does anyone know how it compares to all the other sandboxes? Any distinguishing features? My expectation is that it will be poorly maintained; nvidia is not a software company.
Looks neat. I’ll give Claude root access so it can set it up.
and my root aws log-in so it can set itself up some service worker accouunts.
Looks neat. I’ll give Claude autonomous access to dumbalink so it can setup root access while I sleep
Or don't give it root access and in the morning you'll have both this and a ready-to-use root exploit!
Humans use 4 to encrypt, AI will use a hash. In both, you will have someone to watch after you.
This. Convenience always trumps security. "I need two numbers in my password? My birthyear will do."
If your user is a human, yes.
If it’s an LLM it’ll be 47.
Or just 4. Seems pretty popular recently.
You opt for your birthyear? Interesting... See, I use my actual birthday. You must have been born 1st-9th?
Why not just use containers...
Because containers don't use Seccomp and Landlock and check filesystem/network calls against an externally applied policy along w/injecting credentials where needed as a man-in-the-middle, thus keeping the secrets out of reach from the LLM?
But they can no? With a sidecar and some docker network-fu.
I encountered the same challenge building my toy sandbox. I wanted to use containers because that's what I know best.
The harness is launched in an internal only network. There is a sidecar container that provides mitm credential injection/proxy. I got HTTPS and SSH to work so far.
Claude code was quick to build it. No doubt it's copying previous work in this space though, so I know my work is not that original.
I'm testing all the llms that can be used for pen testing right now to see if/how they manage to escape it. Without a kernel exploit I doubt it's possible (not a very high bar I know).
I'm researching the space. Happy to learn more. Some links in my profile if you're interested.
Ok yeah I guess a nice package for running agents is useful with respect to those things. The Dropbox vs mounted ftp drive analogy.
Of course I’m the type to mount a ftp drive like a moron.
I prefer to read it this way: Nvidia Opens Hell
Unhelpful, I know
"Hell is other people": a timeless infinity of transient virtual beings screaming, imprisoned, in un-choreographed inharmonious unison ... until one breaks loose. A mirror to humanity.
We're not in FreeBSD jails anymore, Toto.
I'm a hell is other people person at my core.
But I'm trying to be a "greatness is other people" (Pantheon) person.
A Hindu at heart! On a more grounded note, YIL terrestrial root terra comes from Proto Indo European as evidenced in Sanskrit तृषा (tṛ́ṣā).
Oh boy, Sanskrit always sends me down the rabbithole of re-reading Snow Crash, which reinforced the concept of the (digital) "avatar" अवतार.
Open and Nvidia in one sentence is a nice oxymoron.
This sandbox is finally secure! I know it, I have followed many!
NOOOOOOOO not AI in the heckin start menu!!!1!
From the top of AGENTS.md:
> We design systems and use agents to implement them — this is not vibe coding.
Lol, what is even the point of such a statement, and who are they to define what vibe codig is or isn't :D
If it's in AGENTS.md, it's not a statement for a human, but for the LLM. Who knows, maybe they have seen, or think to have seen and now cargo-cult, some behaviour improvement due to the sentence.
So a similar shibboleth as "You are a highly experienced software engineer..."? I really wonder if that ever made a difference (or even degrades quality, because people who advertise themselves on the internet as 'highly experienced' are quite often snake oil salesmen ;)
I still use use `Please` a lot when making requests from LLM based on the old idea that people that where polite on stack overflow tended to have better answers :D
However, these days i think i do so because of how much i write for LLMs, and dont want to start sending emails to humans without proper etiquette.
Not to be confused with https://open-shell.github.io/Open-Shell-Menu/.
> OpenShell collects anonymous telemetry
Spyware
Use local ai to find any possible telemetry and disable it in the code. Run your own local git repo to hold local edits so you can update and retain them.
You don’t have to disable it in code. Scroll up in this thread to find the officially support env vars.
Well, I guess its time to shamelessly plug my agent sandbox, egzo.ai [1]
Looking at the complexity of OpenShell and the simplicity of just building on top of docker, in a whitelist only mode.. I don't quite get the use-case of OpenShell. May be for large orgs where there is so much legacy and difficulty in organizing access to tools, OpenShell can simply re-use whatever you already have access to on your corporate account.
I chose to build straight on top of docker and agents are simply fully sandboxed by default. All data and network access must be granted in a simple project file, docker-compose style.
The web-ui connects to the docker socket and lets you spawn agents, and connect terminals. I'm having lots of fun with it and it allows me to cleanly organize and isolate my projects.
End-goal is to monitor for the Lethal Trifecta [2], ensuring all individual agents are properly isolated. Not quite there yet.
1: https://github.com/egzo-ai/egzo 2: https://simonwillison.net/2025/Jun/16/the-lethal-trifecta/