agent-sandbox¶

Self-hostable, disposable sandboxes where AI agents run code, with snapshot, rollback, and fork.

You give an agent a sandbox. The agent writes files, runs commands, and streams the output. Before it tries something risky, it takes a snapshot. When the attempt goes wrong, it rolls back in one call, or forks the snapshot into several sandboxes and tries three ideas in parallel. When the agent forgets to clean up, the TTL reaper does it for you.

agent-sandbox is a small Python server with a REST API, a Python SDK, and an MCP server. It runs on any Linux host with Docker, and your code never leaves your infrastructure.

A terminal session that creates a sandbox, runs code, snapshots, breaks the workspace, rolls back, and destroys the sandbox

Where to go next¶

Quickstart: install the server and run your first command in a sandbox.
Concepts: sandboxes, snapshots, rollback, fork, and TTLs.
Architecture: how the API, the manager, and the backends fit together.
REST API, Python SDK, and CLI reference.
FAQ and the security model.

Install in one line¶

curl -fsSL https://raw.githubusercontent.com/superintelligenceco/agent-sandbox/main/install.sh | sh

The script downloads the standalone agent-sandbox executable for your OS and CPU, checks it against the release's SHA256SUMS, and installs it into ~/.local/bin. The Quickstart covers the Docker image, the Compose file, and the PyPI package.