Parallel agents
Run multiple agents side by side, each with its own session, model, and permissions.
In hob, agent sessions are the primary unit of work. You can run multiple agents in parallel, each in its own pane, working on different parts of your project at the same time.
Launching agents
Press Alt+A to open a new agent pane. Each pane is fully independent — its own conversation, model, backend, and permission settings.
A typical setup might look like:
- One agent refactoring backend logic
- Another writing frontend tests
- A terminal running the dev server alongside
All visible at once in a tiled layout. No hidden tabs.
Configuring each agent
Each agent pane has its own settings in the pane header:
- Backend — choose between Claude Code, Codex CLI, OpenCode, or any configured backend
- Model — pick the model for that session (e.g. Sonnet 4.6, Opus 4.6)
- Permissions — control what the agent can do: ask before each action, approve all, or deny all
You can run different backends in different panes. One agent on Claude Code, another on Codex — whatever fits the task.
Navigating between agents
| Shortcut | Action |
|---|---|
Alt+Arrow | Move focus to adjacent pane |
Alt+Shift+Arrow | Swap pane positions |
Alt+Shift+E | Equalize all pane sizes |
Alt+A | Open new agent pane |
Cmd+W | Close current pane |
Workspaces
Group related panes into workspaces — one workspace per task or feature branch. Switch your entire screen with Ctrl+Tab or Alt+1 through Alt+9.
Each workspace remembers its layout, so you can jump between contexts without losing your place.