NOFire AI today released Brig as an Apache 2.0 Open Source project to run AI coding agents inside a secure and low-overhead microVM on Mac or Linux.
The security concern Brig addresses, is that AI agents can easily install packages, execute any arbitrary action, use the network, and developer's credentials; making it so that infiltration attacks (ie. “Axios supply chain attack”) can occur. Brig closes this gap on both local development machines on MacOS and Linux, as well as Cloud hosts.
As engineers nowadays use auto-approval to accelerate their AI development experience, the whole operating system is at risk. Brig on the other hand, enables them to contain that risk within the boundaries of an ephemeral microVM.This is enforced via hardware-level virtualisation, where memory and process boundaries are physically enforced by the underlying CPU architecture.
You can start with just $ brig run claude
Open Source first
All components are Apache 2.0. Free and Open.
The microVMM code that a security aware enterprise would have to audit is less than 20k lines of code. Making it one of the lowest overhead sandbox technologies in the world, open for audit by any security researcher.
Available Today
Brig is available today, via:
$ curl -fsSL https://brig.sh/install | sh
Execute on MacOS on Apple Silicon or Linux on x86_64 and ARM, using curated profiles for Claude Code, Codex, Cursor, Gemini, Grok and opencode, or bring your own using OCI images (i.e. Ubuntu).
Documentation, the security model and the features of Brig are at brig.sh and in the repository.
From the creators
“Having led the development of the urunc CNCF (Cloud Native Compute Foundation) Sandbox project, focusing on unikernel technologies and microVMM for the majority of my career. The VM/microVM technologies I’ve been working on since 2007, are now more relevant as ever; due to the rise of AI Agents and agentic workloads, require the security and trust boundary to move lower without paying a significant performance tax.”
Anastasios Nanos, Co-founder, Chief Scientist
“NOFire AI is building the Context and Control Model for Production: a live, time-versioned map of how production actually behaves, with governance on every agentic action. In our mission to deliver controlled autonomy for remediation in production systems in a reasonably secure manner, we invested heavily in a technology that is now applicable to so much more.”
Moustafellos Panagiotis, Co-founder, CTO.
Press contact
[email protected]