Back to directory

ipybox

A unified, sandboxable execution environment that runs Python code, shell commands and programmatic MCP tool calls together in a stateful IPython kernel.

ipybox

ipybox executes code blocks containing any mix of Python code, shell commands and programmatic MCP tool calls inside a stateful IPython kernel. Kernels can be sandboxed with filesystem and network restrictions enforced at the OS level, and it generates Python APIs for MCP server tools automatically, so those tools can be called like ordinary functions from code.

It supports application-level approval of individual tool calls and shell commands during execution, and runs locally, giving code protected access to local data and tools rather than routing everything through a remote sandbox. This is a complete rewrite; the previous generation is maintained on an older branch.

It is a general-purpose building block: the Freeact agent is one example of a general-purpose agent built on top of it.

More in Containers

See all 120
  • Drop

    droprun.sh

    A rootless Linux sandbox for developers, built on user namespaces with optional gVisor isolation, that confines coding agents and third-party packages without leaving your normal work environment.

  • minienv

    github.com

    Reads an existing docker compose file and deploys it as a remote preview environment with little setup, with optional public endpoints and authentication so others can review in-flight work.

  • Rig

    npmjs.com

    Gives each coding agent, such as Claude Code, Cursor or Codex, its own cloud Linux desktop with a running dev server, tests and a signed-in Chrome, so the developer's laptop stays free while several agents work at once.