Authentication
AgentDock never handles authentication itself — you authenticate each CLI independently.
AgentDock does not provide accounts, subscriptions, or authentication of its own for any agent. It is a local desktop shell over CLIs you install and authenticate yourself.
What this means in practice
- You authenticate
claude,codex, andagydirectly with their own providers, outside AgentDock — following each provider's own login flow (typically run from that CLI directly in a terminal). - AgentDock never stores, transmits, or manages credentials for any agent. It only detects whether a working, authenticated CLI is available on your system and uses it as-is.
- Usage limits and billing are entirely between you and each agent's own provider — AgentDock has no visibility into or control over them.
Important
If an agent works fine from your regular terminal but fails inside AgentDock, authentication is rarely the cause — check agent CLI not detected first, since AgentDock may simply be launching a different executable (or a different environment) than your terminal does.
Where authentication problems show up
AgentDock surfaces authentication failures as they're reported by each CLI:
- Antigravity's terminal output is pattern-matched for phrases like "not authenticated," "please log in," "run /login," and "you are currently not signed in."
- Claude Code and Codex report authentication failures through their own SDKs, surfaced in the conversation as an error or interaction card.
See Troubleshooting: authentication problems for what to do when this happens.