Skip to main content
Aether binds a handful of concepts together by a single rule: one task = one agent = one workspace VM. Everything below hangs off that.

Projects

The top-level container for your code, environment, GitHub link, and per-project task defaults.

Workspaces

The isolated Linux VM where a task’s agent reads, runs, and tests your code.

Tasks & Lifecycle

A unit of agent work, bound to one workspace, that moves through queued → processing → awaiting input.

Agents & Models

The three selectable providers — Codex (default), Claude Code, OpenCode — and the model each runs.

Billing & Credits

VM runtime is metered in credits. Model calls on Aether inference are charged to a separate prepaid balance; tokens never affect credits.

Integrations

GitHub, Sentry, Linear, and Slack connections that let the agent work with your repos, errors, issues, and notifications.

​
How they fit together

You create a project (optionally linked to a GitHub repo) and start a task with a prompt. Aether provisions one isolated workspace VM and runs one agent inside it. The agent works until it finishes a turn or needs your input; you respond and it resumes. The VM’s runtime is metered in credits. Integrations let the agent pull context (issues, errors) and push results (pull requests, notifications) to the tools you already use.