This looks really good - the CLI interface design is solid, and I especially like the secrets / network proxy pattern - but the thing it needs most is copiously detailed documentation about exactly how the sandbox mechanism works - and how it was tested.

There are dozens of projects like this emerging right now. They all share the same challenge: establishing credibility.

I'm loathe to spend time evaluating them unless I've seen robust evidence that the architecture is well thought through and the tool has been extensively tested already.

My ideal sandbox is one that's been used by hundreds of people in a high-stakes environment already. That's a tall order, but if I'm going to spend time evaluating one the next best thing is documentation that teaches me something about sandboxing and demonstrates to me how competent and thorough the process of building this one has been.

UPDATE: On further inspection there's a lot that I like about this one. The CLI design is neat, it builds on a strong underlying library (the OpenAI Codex implementation) and the features it does add - mainly the network proxy being able to modify headers to inject secrets - are genuinely great ideas.

> There are dozens of projects like this emerging right now. They all share the same challenge: establishing credibility.

Care to elaborate on the kind of "credibility" to be established here? All these bazillion sandboxing tools use the same underlying frameworks for isolation (e.g., ebpf, landlock, VMs, cgroups, namespaces) that are already credible.

The problem is that those underlying frameworks can very easily be misconfigured. I need to know that the higher level sandboxing tools were written by people with a deep understanding of the primitives that they are building on, and a very robust approach to testing that their assumptions hold and they don't have any bugs in their layer that affect the security of the overall system.

Most people are building on top of Apple's sandbox-exec which is itself almost entirely undocumented!

> The problem is that those underlying frameworks can very easily be misconfigured.

Agreed. I'm sure a number of these sandboxing solutions are vibe-coded, which makes your concerns regarding misconfigurations even more relevant.

I'm sure 100% of them are vibe coded. We were all wondering where this new era of software is, and now it's here, a bunch of nominally different tools that all claim to do the same thing.

I'm thinking the LocalLLM crowd should take their LLMs to trying to demolish these sandboxes.

Simon! Thanks. I appreciate your comment and totally agreed. I will improve the docs as well as tests.

We've seen cases where filesystem sandboxing alone wasn't enough — outbound requests still allowed credential exfiltration attempts.

Compare with and steal any ideas you like from mine if you like. I've got a semi-decent curl|bash pattern covered, and also add network filtering via pasta (which may be more robust than rolling your own). https://github.com/reubenfirmin/bubblewrap-tui

Ohh! thanks for sharing this. You are using DNS proxy which is interesting and useful if a process doesn't respect the HTTPS_PROXY/HTTP_PROXY/etc. env vars that I'm injecting. I will take a look, very interesting.

You should probably add a huge disclaimer that this is an untested, experimental project.

Related, a direct comparison to other sandboxes and what you offer over those would be nice

I agree to some extend. I'm using the OpenAI Codex crates for sandboxing though, which I think it's properly tested? They launched last year and iterated many times. I will add a note though, thanks!

This is more a criticism of codex's linux-sandboxing, which you're just wrapping, but it's the first I've ever looked at it. I don't see how it makes sense to invoke bwrap as a forked subprocess. Bubblewrap can't do anything beyond what you can do with unshare directly, which you can simply invoke as a system call without needing to spawn a subprocess or requiring the user to have bwrap installed. It kinds of reeks of amateur hour when developers effectively just translate shell scripts into compiled languages by using whatever variant of "system" is available to make the same command invocations you would make through a shell, as opposed to actually using the system call API. Especially when the invocation is crafted from user input, there's a long history of exploits arising from stuff like this. Writing it in Rust does nothing for you when you're just using Rust to call a different CLI tool that isn't written in Rust.

Is your criticism here that there's no point in invoking bwrap directly when you could instead implement the same things that bwrap implements?

I'd much rather a system call bwrap than re-implement bwrap, because bwrap has already been extensively tested.

That was my thinking, too. The only other option would be reimplement it in Rust (never researched what exists though).

Thanks for sharing this, I read your comment multiple times. What would be the alternative though? It is true that the program being written in Rust doesn't solve the problem of spawning subprocesses, but what's the alternative in that case?

Oh wow, this looks nicely done! It's also nice that it's cross platform. I've done something similar with https://github.com/Gerharddc/litterbox which takes things a bit further by allowing you to easily sandbox your entire development environment (i.e. IDE and everything) using containers. Unfortunately I have not gotten around to the network sandboxing part though, that seems very tricky to get useful without being too "annoying".

Thanks for sharing this. I really like the idea

Hey - I'd love for you to add a documented / standard way to use this inside dockers so we can use build on it for various agentic efforts. I've solved getting bubblewrap to work inside a docker once for the nanobot project, but the folks there are dragging their feet on incorporating sandboxing.

https://github.com/HKUDS/nanobot/pull/1940

I've been testing this on Docker today, including the credential injection, env vars, net calls control. I will add more docs but one interesting use case would be to have something like `zerobox --profile nanoclaw -- nanoclaw`, or something similar.

I'd like to hear your thoughts.

I'll give it a shot later today, but basically you need a pretty specific seccomp profile (see my example - I pulled from the podman repo) to allow bubblewrap to run inside an unpriviledged docker.

This is really useful! How well does it compare though to Docker etc.

Because I am worried about sandbox escapes. This is what we currently use to sandbox JS inside Browsers and Node (without anything extra) : https://github.com/Qbix/Platform/blob/main/platform/plugins/...

I like tools like this, but they all seem to share the same underlying shape: take an arbitrary process and try to restrict it with OS primitives + some policy layer (flags, proxies, etc).

That works, but it also means correctness depends heavily on configuration, i.e. you’re starting with a lot of ambient authority and trying to subtract from it enforcement ends up split across multiple layers (kernel, wrapper, proxy)

An alternative model is to flip it: Instead of sandboxing arbitrary programs, run workflows in an environment where there is no general network/filesystem access at all, and every external interaction has to go through explicit capabilities.

In that setup, there’s nothing to "block" because the dangerous primitives aren’t exposed, execution can be deterministic/replayable, so you can actually audit behavior. Thus, secrets don’t enter the execution context, they’re only used at the boundary

It feels closer to capability-based systems than traditional sandboxing. Curious how people here think about that tradeoff vs OS-level sandbox + proxy approaches.

Zerobox uses the same kernel mechanisms (namespaces + seccomp) but no daemon, no root and cold start ~10ms (Docker is much worse in that regard).

Docker gives you full filesystem isolation and resource limits. Zerobox gives you granular file/network/credential controls with near zero overhead. You can in fact use Zerobox _inside_ Docker (e.g. for secret management)

Personally I would probably always reach for a docker container if I want a sandboxed command that can run identically anywhere.

I appreciate that alternate sandboxing tools can reduce some of the heavier parts of docker though (i.e. building or downloading the correct image)

How would you compare this tool to say bubblewrap https://github.com/containers/

The text says that it uses OS-level tools, specifically bubble wrap on Linux.

That's right. It uses the same kernel mechanisms as Docker, the runtime is different though (bwrap on linux, seatbelt on mac, etc.)

Very interesting. I just started researching this topic yesterday to build something for adjacent use cases (sandboxing LLM authored programs). My initial prototype is using a wasm based sandbox, but I want something more robust and flexible.

Some of my use cases are very latency sensitive. What sort of overhead are you seeing?

I added a benchmark test (Apple M5) and on average I'm seeing 10ms overhead. I added a benchmark section to the repo as well https://github.com/afshinm/zerobox?tab=readme-ov-file#perfor...

Also, I'm literally wrapping Claude with zerobox now! No latency issues at all.

Wasm sandboxes are fast for pure compute but get painful the moment LLM code needs filesystem access or subprocess spawning. And it will, constantly. Containers with seccomp filters give you near-native speed and way broader syscall support — overhead is basically startup time (~2s cold, sub-second warm). For anything IO-heavy it's not even close. We're doing throwaway containers at https://cyqle.in if anyone's curious.

I will run the same benchmark test on wasm sandboxes just to be able to compare it with Zerobox. I will share the results tomorrow.

Here is the video, running Claude with Zerobox, you can see the latency, etc. https://www.youtube.com/watch?v=xzsGsSsx0OI

It’s terrific to see this. I’m definitely going to give it a whirl. I’ve been working on a specific JavaScript isolate[^1]. This is great source of inspiration for it.

[^1]: https://github.com/jonathannen/hermit

I'd love to hear your thoughts! I've been primarily testing this with Bun + Vercel AI SDK for tool call sandboxing.

> zerobox --secret OPENAI_API_KEY=$OPENAI_API_KEY

Linux by default allows all users to read CLI arguments of running processes. While it looks like your bwrap invocation prevents the sandbox from looking at this process (--unshare-pid), any other process running on your system can read the secret.

That's true and the expected behaviour but I see your point. The example there is not great, I should've used `sk_s123...` to show that you are passing the env var to the sandbox as opposed to setting it on the host, then proxying it. I will update it.

Again, it’s blacklisting so kind of impossible to get right. I’ve looked at this many times, but in order for things to properly work, you have to create a huge, huge, huge, huge sandbox file.

Especially for your application that you any kind of Apple framework.

This doesn't look like it's blacklisting to me. It's an allowlist system:

  --allow-net=api.openai.com # Explicitly allow access to that host

  --allow-write=config.txt # Explicitly allow write to that file

That's correct. The pattern is: reads allowed, write and network I/O blocked by default.

```

zerobox -- curl https://example.com

Could not resolve host: example.com

```

Oh so it allows ALL file reads?

I'd feel safer with default-deny on reads as well, but I know from past experience that this gets tricky fast - tools like Node.js and uv and Python all have a bunch of files they need to be able to read that you might not predict in advance.

Might still be possible to do that in a DX-friendly way though, if you make it easy to manually approve reads the first time and use that to build a profile that can be reused on subsequent command invocations.

I agree and you can deny all reads like this:

```

zerobox --deny-read=/ -- cat /etc/passwd

```

That being said, what the default DX shouldl be? What paths to deny by default? That's something I've been thinking about and I'd love to hear your thoughts.

That's a really tough question. I always worry about credentials that are tucked away in ~/.folders in my home directory like in ~/.aws - but you HAVE to provide access to some of those like ~/.claude because otherwise Claude Code won't work.

That's why rather than a default set I'm interested in an option where I get to approve things on first run - maybe something like this:

  zerobox --build-profile claude-profile.txt -- claude

The above command would create an empty claude-profile.txt file and then give me a bunch of interactive prompts every time Claude tried to access a file, maybe something like:

  claude wants to read ~/.claude/config.txt
  A) allow that file, D) allow full ~/.claude directory, X) exit

You would then clatter through a bunch of those the first time you run Claude and your decisions would be written to claude-profile.txt - then once that file exists you can start Claude in the future like this:

  zerobox --profile claude-profile.txt -- claude

(This is literally the first design I came up with after 30s of thought, I'm certain you could do much better.)

Fantastic! I like that idea. I'm also exploring an option to define profiles, but also have predefines profiles that ships with the binary (e.g. Claude, then block all `.env` reads, etc.)

Being able to mix and match profiles would be neat.

Give me 2 days :)

The `--build-profile` / `--profile` thing is a good idea, but typically you'd want to just save all of the access that the program does without prompting.

Programs will access many files and directories on startup, and it would be extremely tedious to have to manually approve each one. So you'd auto-approve all and save them to the profile. This is TOFU principles applied to sandboxing. The assumption being that "this first time I run it naked, it's unlikely to do anything malicious, let me enforce that behavior for the future."

I agree. What would be the ideal DX from your point of view?

The DX above from @simonw seems perfectly fine.

Let the user play with the app and after they exit the profile should contain all of the access attempts in a human readable format that's editable by the developer.

There might be many access attempts to folders in one directory, e.g.:

~/Documents/...

So instead of having a massive list of files it should be easy for developers to edit the profile to say, "Allow everything there", e.g. ~/Documents/*

That's interesting, thanks for sharing that. Could you elaborate a bit more? I'd like to understand the use case is a bit better.

I trust sandbox-exec more, or Docker on Linux. Those come from the OS, well tested and known.

MITM proxy is nice idea to avoid leaking secrets. Isn’t it very brittle though? Anthropic changes some URL-s and it’ll break.

Thanks for sharing that. Zerobox _does_ use the native OS sandboxing mechanisms (e.g. seatbelt) under the hood. I'm not trying to reinvent the wheel when it comes to sandboxing.

Re the URLs, I agree, that's why I added wildcard support, e.g. `*.openai.com` for secret injection as well as network call filtering.

You know, the thing is, that it is super easy to create such tools with AI nowadays. …and if you create your own, you can avoid these unnecessary abstractions. You get exactly what you want.

How do you intercept network traffic on mac os? How do you fake certificates?

Zerobox creates a cert in `~/.zerobox/cert` on the first proxy run and reuses that. The MTIM process uses that cert to make the calls, inject certs, etc. This is actually done by the underlying Codex crate.

Yeah, but how does the sandboxed process “know” that it has to go through the proxy? How does it trust your certificate? Is the proxy fully transparent?

Oh I see. It inject HTTP_PROXY/HTTPS_PROXY/etc. env vars into the process so that all sandboxed subprocesses go through the proxy.

What if the program doesn’t respect those env vars? Can Zerobox still block network calls in that case?

Great question! On Linux, yes, network namespaces enforce that and all net traffic goes through the proxy. Direct connections are blocked at the kernel level even if the program ignores proxy env vars, but I will test this case a bit more (unsure how to though, most network calls would respect HTTPS_PROXY and other similar env vars).

That being said, the default behaviour is no network, so nothing will be routed if it's not allowed regardless of whether the sandboxed process respects env vars or not.

Does this work inside of Podman containers?

How about on macOS?

On macOS, the proxy is best effort. Programs that ignore HTTPS_PROXY/HTTP_PROXY can connect directly. This is a platform limitation (macOS Seatbelt doesn't support forced proxy routing).

BUT, the default behaviour (no net) is fully enforced at the kernel level. Domain filtering relies on the program respecting proxy env vars.

I thought seatbelt-exec had mechanisms for that?

  (allow network-outbound
    (remote tcp "127.0.0.1:8080"))

It does but because I'm inheriting the seatbelt settings from Codex, I'm not resetting it in Zerobox (I thought it's a safer option). Let me look into this, there should be a way to take Codex' profile and safely combine/modify it.

Cool project! I think there would be a lot of value in just having a mode that logs all the file operations a script tries to make. Great work!

Cool project, and I think there would be a lot of value in just logging all operations.

For just logging would it really give any more info than a trace already does?

Forgot about that, was mostly thinking about how AI agents with unrestricted permissions would ideally have some external logging and monitoring, so there would be a record of what it touched. A trace has all of the raw information, so some kind of wrapper around that would be useful.

I'd like to know what level of details you'd expect. Something like `zerobox -- claude`, then you get an output log like this:

```

Read file /etc/passwd

Made network call to httpbin.org

Write file /tmp/access

```

etc.? I'm really interested to hear your thoughts and I will add that feature (I need something like that, too).

*strace that is - annoyingly it seems it was autocorrected away

I think there is still a valid case for sandbox logs/otel. strace would give you the syscalls/traces but not _why_ a particular call was blocked in side the sandbox (e.g. the decision making bit).

Agreed. I added the `--debug` flag this morning. It does simple logging including the proxy calls:

```

$ zerobox --debug --allow-net=httpbin.org -- curl

2026-04-01T18:06:33.928486Z CONNECT blocked (client=127.0.0.1:59225, host=example.com, reason=not_allowed)

curl: (56) CONNECT tunnel failed, response 403

```

I'm planning on adding otel integration as well.

Very cool. Is there a way to have a notion of a session, saving state between runs?

No, it's stateless right now. What is your requirement though? How do you define a session? Are you referring to "snapshotting" between sessions?

I'm adding snapshotting as well https://github.com/afshinm/zerobox/pull/21

Then you can run:

```

zerobox --snapshot -- sh -c 'echo "abc" > a'

```

and also `zerobox snapshot list/diff/restore`

there's been so many of these -- which of these sandboxing tools is best?

What's the right problem to be solving here?

I'd love to learn more please. I'm interested in sandboxing AI tools/agents regardless of the underlying mechanism (I explored Firecracker VMs briefly as well, terrible cross platform support though).

Does Zerobox support audit logging for blocked network or file operations?

I added some basic --debug support earlier today, but I will work on proper JSONL/Otel integration soon.

Wish it wasn’t rust… it’s so hard to read.

I know. I will add more docs soon though, that should make it easier to navigate the code and understand what's going on.

I love sandboxes man