Configurable sandboxes for long-running AI agents

Set up your environment once. Fork it for every task.

How it works

  1. Configure — Set up a workspace with your repos, dependencies, and AI harness (Claude Code, Amp, Codex, etc.)
  2. Launch — Start isolated tasks with a prompt. Each task runs in its own sandbox.
  3. Monitor — Connect via terminal, web, or phone to monitor and interact with your harness, or running services.

Features

  • Fully configurable — Install packages, clone repos, and set up your agent harness exactly how you want it with an interactive setup.
  • Background services — Run databases, dev servers, or any long-running process. They restart automatically on wake.
  • Expose ports — Make services accessible via public URL.
  • Launch from anywhere — Start tasks from the CLI, API, Web or your phone.
  • SSH access — Connect directly to any running sandbox.