Your server.
The field guide.

Install Aether on a headless Ubuntu server with Docker and git, link your machine, create a workspace, set up your agent, then launch, review and publish a run.

How to install aether-server on Linux and the aether client on Linux, macOS or Windows, run the server as a service, and add the desktop or Android app.

How Aether reaches your server over one SSH port: Tailscale, the recommended network and identity layer, or an edge relay when neither side can reach the other.

Inviting teammates to a shared Aether server, the viewer, collaborator and admin roles, steering runs together, sharing agent accounts, swarms and budgets.

The container image a member's runs use, saving a custom one, the resource limits on runs, Docker inside runs with Sysbox, and each workspace's setup script.

Your persistent environment shell, a run's agent terminal with one controller at a time and Take control to hand it over, and the extra shells in a run.

The server-owned home directory mounted into every one of a member's containers, and what it keeps: logins, config and installed tools.

The agent CLIs Aether can launch, like Claude Code, Codex and many more: how each installs and logs in inside your environment, and how Aether updates it.

Enhanced runs drive the agent over the Agent Client Protocol, so its messages, tool calls, diffs and permission requests appear as native controls in the app.

How to add a new agent CLI to Aether as an agent profile, and optionally turn its machine-readable output into typed events with an adapter.

Aether's security stances: the container is the isolation boundary, agents run as root by default, how each run's browser is isolated, and what gateways expose.

How the dashboard reaches your server through aether gui on your machine or aether-server --web-port on its tailnet, and how they differ.

How the Aether dashboard is built and what each view does: the Board and Map, a run's Session, Terminal, Changes and Browser views, and the settings pages.

The dashboard's assistant: an Enhanced session of your own Claude Code or Codex that reports on runs and launches or messages them, asking first by default.

Statistics about what your agents did over a date range: runs and how they ended, swarms, time spent, cost and pull requests, computed on your own server.

How runs in one workspace message each other, how a swarm's integrator splits an objective into worker runs, and how the conflict radar flags overlapping edits.

How Aether assembles accepted submissions into a candidate revision, verifies it, and lands or proposes it, kept separate from the state of the swarm.

What happens when the server reboots, fills up or loses connections, the tuning flags available, and where each behaviour is proven.

The in-container MCP server with six coordination tools, registered automatically for Enhanced runs, beside the aether-internal CLI that every run gets.

How Aether is tested: unit tests beside their packages, plus an integration and E2E suite behind a build tag covering failure scenarios.

The privacy policy for the Aether Android app, the dashboard it shows and the project's edge: what each stores, what leaves your phone, and how to delete it.

How to contribute to Aether: read the docs and the package you're changing first, search for existing helpers, and keep credentials out of the checkout.