Orca CLI

The orca CLI scripts the entire app: create worktrees, run terminals, drive the embedded browser, take snapshots, and send prompts — every workflow you click, as a composable command.

How agents use it

The Orca CLI is how agents drive the app back. An agent with shell access can:

# create a worktree and run in it
orca worktree create --name fix-api --agent claude \
  --prompt "Fix the API timeout"

# drive the embedded browser
orca snapshot            # accessibility tree of the active page
orca click --ref btn-save
orca fill --ref input-email "test@example.com"

# notify the human
orca notify --title "Agent done" --body "PR ready for review"

Your agents don’t just run in the app — they operate it, turning the IDE itself into their tool surface.

Common commands

Command	What it does
orca worktree create	New isolated worktree + agent
orca snapshot	Accessibility tree of the active browser page
orca click / orca fill	Browser interaction by element ref
orca terminal list / send	Watch and drive terminals
orca notify	Desktop notification
orca tab create / close	Embedded browser tabs

Scripting patterns

• Fan out: loop worktree create across N agents, then diff the worktrees.
• Verify: snapshot → assert the expected UI state → notify with the result.
• Hand off: agent A creates the worktree; agent B (a different CLI) works in it; agent C reviews.

Tips

• Every command supports --json for scriptable output.
• orca --help lists everything; each subcommand has its own --help.
• The CLI talks to the running app — start Wakii first.