getkern/kern

Rootless container runtime with no daemon. Runs agent tool calls and LLM-generated code in kernel-enforced containers.

What it solves

Kern provides a lightweight, daemonless container runtime specifically designed for running untrusted code generated by AI agents. It addresses the need for secure, isolated execution environments that can be spun up in milliseconds, allowing agents to execute code, run tools, or manage services without risking the host machine.

How it works

Kern operates as a single static binary that manages OCI-compliant containers using kernel-level enforcement. It utilizes Linux namespaces, cgroups v2, and a strict seccomp allowlist to sandbox processes. Because it has no background daemon, it consumes zero memory when idle. It provides a dedicated SDK (for Python and Node) and an MCP (Model Context Protocol) server, enabling AI agents in tools like Cursor, Claude Code, or LM Studio to trigger code execution directly within these isolated containers.

Who it’s for

This tool is for developers building AI agents or LLM-powered applications who need a secure way to execute model-generated code or manage tool-use environments. It is also suitable for users who want a faster, daemonless alternative to traditional container runtimes like Docker or Podman for local development.

Highlights

  • Daemonless Architecture: Runs as a single static binary with no background processes or sockets, resulting in zero memory usage at rest.
  • High Performance: Starts containers in single-digit milliseconds, significantly faster than traditional container runtimes.
  • Agent-Ready: Includes native support for the Model Context Protocol (MCP) and SDKs for Python and Node to facilitate seamless integration with AI agents.
  • Security-First: Employs kernel-enforced boundaries, including a strict seccomp allowlist, read-only root filesystems, and dropped capabilities to safely run untrusted code.
  • Compatibility: Supports standard OCI images and can run existing docker-compose.yml files.

Related

  • Dispatch
    Kern v0.7.0: Daemonless, Rootless Container Runtime in a 1.5 MB Binary

    Kern v0.7.0 is a fast, daemonless, and rootless sandbox and virtual resource runtime that can start OCI images in ~3.5 ms, packaged as a single 1.52 MB static binary.

  • Project
    multikernel/sandlock

    A lightweight, rootless Linux sandbox for isolating untrusted code using Landlock and seccomp, offering high-performance security for AI agents and scripts without the overhead of containers.

  • Project
    vndee/llm-sandbox

    A secure execution environment for LLM-generated code that supports multiple container and local OS backends to safely run and manage code across various programming languages.

  • Project
    superradcompany/microsandbox

    A tool for running untrusted workloads in fast, local microVMs with hardware-level isolation, specifically designed for AI agents and automated code execution.

  • Project
    open-webui/open-terminal

    Open Terminal is a self‑hosted service that exposes a remote shell, file browser and code‑execution via a REST API, letting AI assistants run commands safely. It can be run in Docker (sandboxed with many pre‑installed tools) or installed as a Python package for direct host execution, supports configurable images, runtime package installation, optional multi‑user workspaces, and integrates with Open WebUI.