Skip to content
Aether Platform

Agent Runtimes

One of the core functions of the AI Namespace is the Agent Runtime, designed to safely execute autonomous AI agents (like Devin, Cursor, AutoGPT).

Sandboxed Environment

To ensure that code and commands generated by agents do not affect the host system, Aether provides a highly isolated sandbox.

  • Secure Runtime Isolation: We use security-hardened runtimes to prevent container escapes and ensure host safety.
  • Ephemeral Filesystem: The filesystem manipulated by the agent is temporary and completely reset after task completion.

Tool Access Integration

Agents are not isolated from utility. Aether’s runtime provides standard interfaces for agents to access key tools:

  • Terminal Access: Execute commands via a secure shell (e.g., npm install, python script.py).
  • Browser Access: Retrieve web page information via headless browsers.
  • File Edit: Read/Write permissions for specific files within the workspace.

Long-running Tasks

Even if the IDE is closed, agents can continue tasks in the background for hours. This is ideal for “tasks humans can’t wait for,” such as large-scale refactoring or automated test generation.