Skip to content

Known limitations

Current gaps worth knowing about before you rely on AgentDock for something they'd affect.

  • Unsigned Windows and macOS builds — no code-signing certificate is configured for either platform yet, so SmartScreen and Gatekeeper warnings on first launch are expected, not a sign of a bad download.
  • No in-app auto-updater — updating means downloading and installing the latest release yourself; see Updating AgentDock.
  • Linux packaging gap between repositories — AgentDock Pro packages Linux as AppImage + deb with native CI; the public free repository has no dedicated Linux packaging script today, and Linux support there remains experimental — see Linux.
  • Provider-specific capability gaps — not every agent supports every feature (live model switching, native image attachments, live permission-mode switching); see Agent capabilities.
  • Compare Mode is read-only isolation, not concurrent-write worktrees — every pane is forced into a safe permission mode rather than given its own isolated writable worktree; see Workspace behaviour.
  • Antigravity's integration is text-classification based — there's no structured protocol backing it, so its prompt/permission detection can be more fragile to upstream CLI changes than the SDK-backed agents.
  • External provider quotas and authentication are entirely outside AgentDock's control — AgentDock has no visibility into your usage limits or billing with Anthropic, OpenAI, or Google.

Related