Multi-Agent Workflows
Running multiple agents in parallel works well in Verne, provided each agent has its own working surface.
Isolate each agent
Open multiple workspaces, or create git worktrees within a single repo. A worktree checks out a separate branch into its own directory, so two agents can write files and run commands without touching each other’s work. Each worktree appears as its own workspace in Verne.
Two agents on the same working tree will overwrite each other. On separate worktrees they work independently and can be diffed once both finish.
Watch the Agents panel
The Agents panel lists every agent across all your open workspaces. Watch the status icons:
| Icon | State | What to do |
|---|---|---|
| Working | Leave it alone for now. | |
| Blocked | The agent is waiting and needs you. | |
| Done | The run finished and is flagged for acknowledgement. |
Scope the panel to a single workspace when you want to focus, or keep it on all workspaces to see the full picture.
Keeping agents focused
Give each agent a clear prompt tied to its worktree’s purpose. If you’re using
Claude Code’s @path mentions,
inject file paths from the editor to steer
it without switching contexts.
Next steps
- Workspaces & worktrees: creating and switching worktrees.
- Isolate work in a worktree: give an agent its own branch to work on.
- The Agents panel: tracking all your agents from one place.