Allox CLI
Allox CLI creates an isolated agent environment, runs shell, browser and Jupyter, saves checkpoints and restores work.
- 163 ★ on the repository
- Last commit
Broad access
Why this level
- The environment runs commands and code, opens a browser and works with files.
- Docker is required and MCP tools inside the environment are called directly.
Install
- Works in
Manual install
uv sync --no-editableAfter cloning the AlloxOrg/allox-cli repository; then start the `opensandbox-server`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The environment runs commands and code, opens a browser and works with files.
- Docker is required and MCP tools inside the environment are called directly.
This is third-party code. Review the repository files before installing.
What it does
The console tool combines an isolated environment, agent tools, file transfer, the current session and checkpoint recovery under one `allox-cli` command. The environment is created through an OpenSandbox server, by default with the AIO Sandbox runtime in Docker. After creation it becomes the current session and later commands need no id. The environment offers a shell, files, a browser, Jupyter, screenshots and MCP: you can discover MCP servers, list their tools and call them directly. Checkpoints are taken manually, on a schedule or after a successful operation and recovery takes `latest` or a chosen version. File transfer is streaming and recursive and symlinks and path escapes are rejected. Output is `table`, `json`, `yaml` or `raw`.
Who it is for. Agent developers who need a reproducible sandbox with resume after a failure.
Good fit when
- An agent runs a long task and checkpoints are needed in case of a break.
- Code must run in an isolated environment with a browser and Jupyter.
- A script or CI needs structured JSON output about the environment.
- You want to see and call the tools of an MCP server inside the sandbox.
Not a fit when
- You have no Docker and no local OpenSandbox server: the environment will not start without them.
- You need a one-command install: currently only from source.
- You need ready cloud environments with no setup of your own.
Example request
Create an allox-cli environment, run a small Python script in it and take a checkpoint.Limitations
The version installs only from source with `uv sync --no-editable`. Docker and a local OpenSandbox server from the `opensandbox-server` package are required and the server is checked with a `/health` request. Do not run `uv run allox-cli` without `--no-editable`: the project would become editable. The README is in Chinese and metadata shows no license file. The project is young, supports one OpenSandbox and AIO Sandbox pair and its stability on third-party images is not described. Checkpoints take disk space, so watch the size with a frequent schedule.
How to disable. Destroy the environment with the `allox-cli sandbox` delete command, stop the OpenSandbox server and delete the project directory and `.venv`.
FAQ
What is needed to run it?
Python 3.10 or newer, `uv`, Docker and a local OpenSandbox server. Docker pulls the AIO Sandbox image itself.
How do I check the server?
With `curl http://127.0.0.1:8080/health` in another terminal after starting `opensandbox-server`.
How do I resume work?
From a checkpoint: `latest` or a chosen version. Checkpoints are made manually, on a schedule or after a successful operation.
Which output formats exist?
`table`, `json`, `yaml` and `raw`, picked for the scenario: a person, a script or CI.
Can I switch environments?
Yes, profiles switch dev, staging, prod, custom images and lightweight runtimes.
Discussion
No questions or reviews yet.
Related
CLI toolsA CLI that runs Claude Code and Codex in isolated Docker containers with a firewall, worktrees and a no-Dockerfile image.
ANOLISA is a system stack for AI agents: output compression, memory, workspace snapshots, command checks and token accounting
A Kubernetes panel with an MCP server: Claude Code reads pod logs, edits resources and rolls back releases in connected clusters.
A container for Claude Code, Codex, Copilot, OpenCode and Pi: the agent runs without prompts while a gateway limits network.
A Docker sandbox for coding agents: a Debian image with Copilot CLI, Codex and Pi, and SSH, RDP and Docker enabled by flags.
Containarium: isolated boxes for AI agents
Containarium
A platform on your own server that gives each agent its own Linux box with SSH and MCP tools for commands and files
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Allox CLI (this card) | no | 163 | ||
| AIO Sandbox | no | key is optional | 6,331 | AIO Sandbox is the container itself; Allox CLI drives it and adds a session, checkpoints and file transfer |
| Agent-Sandbox | no | key is optional | 232 | a self-hosted Kubernetes sandbox with REST and MCP; Allox CLI works locally through OpenSandbox and Docker |
README in short
The Chinese README presents a CLI with checkpoints, a session, an agent toolset, runtime readiness, file transfer, MCP and profiles. It covers install from source, OpenSandbox server setup, end-to-end scenario examples, tests and documentation.