slipstream-inspector

flutter-slipstream

The flutter-slipstream plugin provides MCP tools for launching, observing, and interacting with a running Flutter app. Use these tools — not terminal commands — for all Flutter app work.

Use the plugin tools, not the terminal

When working on a Flutter app with this plugin installed, always prefer the MCP tools over shell commands:

Instead of...	Use...
flutter run	run_app
Hot reload (shell)	reload
Reading terminal log	get_output

run_app does more than flutter run: it auto-selects the best available device, exposes app output via get_output, and enables screenshot capture, layout inspection, and UI interaction — none of which are possible from a plain terminal session.

Standard workflow

run_app → get_output → take_screenshot
  → edit source → reload → get_output → take_screenshot → repeat

• Call get_output after run_app, after reload, and after any interaction tool. It drains the output buffer — skipping it causes output to accumulate silently across calls.
• Call take_screenshot after every reload to confirm the change looked right. Never assume a visual edit was correct without seeing it.
• When get_output returns a [flutter.error] line, the widget ID in the summary can be passed directly to inspect_layout.

Orienting on an unfamiliar screen

1. take_screenshot — see what's on screen.
2. get_route — see the navigator stack and current screen widget names with source file locations. Use before editing to confirm which file to change.
3. get_semantics — see the interactive elements and their node IDs.

Interacting with the app

When slipstream_agent is installed in the Flutter app, prefer finder-based tools — they are more reliable than semantics-based interaction:

• perform_tap(byKey: ...) / perform_set_text(byKey: ...) — most reliable; add ValueKey('my_key') to the widget while writing it.
• perform_tap(byText: ...) — good for button labels.
• perform_tap(byType: ...) — useful for unique widget types.