Skip to main content
Every agent runs under your own provider account, so you must connect a provider before running a task with it — including the default agent, Codex. Each provider has its own connect flow — a device OAuth, an API key, a setup token, or an uploaded auth file. Connections live at the account level and apply to every project. Connect from the web app under Settings → Connections, or with the CLI. Each provider’s CLI commands share the same shape: connect, status, and a disconnect command.

Codex

Codex supports two methods: a ChatGPT-subscription device login, or an API key.
In Settings → Connections, choose Codex and either start the device login or paste an API key.
The device flow stores subscription tokens that Aether refreshes as needed. The underlying endpoints are POST /codex/device-auth/start, POST /codex/device-auth/poll, PUT /codex/auth/api-key, and GET/DELETE /codex/auth — see the Codex auth API.

Claude Code

Claude Code authenticates with a setup token generated by the Claude CLI.
In Settings → Connections, choose Claude Code and paste the setup token printed by claude setup-token.
The token is sent to PUT /claude-code/auth/setup-token — see the Claude Code auth API.

Cursor

Cursor imports the OAuth subscription tokens written by your local Cursor login (cursor-agent login).
In Settings → Connections, choose Cursor and provide your Cursor subscription tokens.
Aether stores both tokens and rotates them as Cursor refreshes them. The import posts to PUT /cursor/auth/subscription — see the Cursor auth API.

OpenCode

OpenCode authenticates with the auth.json file written by your local OpenCode login (opencode auth login). The file is uploaded verbatim and materialized into each task’s config directory at session start.
In Settings → Connections, choose OpenCode and upload your auth.json.
The file is sent to PUT /opencode/auth/file — see the OpenCode auth API.
OpenCode models routed through OpenRouter can additionally use your own OpenRouter API key (aether apikey add --provider openrouter). See Agents and models.

Next Steps

Agents and Models

Pick the agent, model, and reasoning effort per task.

Tasks

Run a task with your connected provider.