mcp-tools.md

MCP tool reference

Back to README · Contributing

Tool	Action
mobile_open_simulator	Open the native panel, optionally targeting a platform, device kind, or discovered device ID
mobile_open_workspace	Open fullscreen with the same optional device targeting
mobile_open_app_intel	Open App Intelligence, optionally focused on an exact screen ID from mobile_app_intel
mobile_list_simulators	Start the bundled backend if needed and list devices
mobile_choose_devices	Ask the user to choose task targets in the host's native form; listed only for hosts with one (Codex, Claude Code)
mobile_list_ios_devices	Discover physical iPhones and iPads with USB/Wi-Fi, pairing state, UDID, and CoreDevice ID
mobile_start_baguette	Retry or reconnect the bundled backend
mobile_boot_simulator	Boot one listed device
mobile_shutdown_simulator	Shut down one listed device
mobile_describe_ui	Read a booted device's accessibility tree for panel annotations, app-only
mobile_repair_input	Reclaim input from Device Hub, closing running apps
mobile_stream_session	Create a panel stream, app-only
mobile_ios_mirror_input	Send physical iOS panel touches for the current video generation, app-only
mobile_stream_input	Send a batch of panel input, app-only
mobile_stream_reset	Recover one iOS panel's MJPEG capture, app-only
mobile_stream_close	Close a panel stream, app-only
mobile_generate_app_intel	Generate or replace a React Native project's static screen and dependency graph
mobile_app_intel	Search the saved graph or query a screen/file's incoming or outgoing relationships
mobile_pick_app_intel_project	Choose a project with the host folder picker, app-only
mobile_select_app_intel_project	Switch to a remembered project, app-only
mobile_app_intel_snapshot	Read the selected project's graph for the visualization, app-only
mobile_app_intel_status	Read generation progress and saved-snapshot status, app-only
mobile_clear_app_intel	Delete the selected project's saved graph, app-only
mobile_cancel_app_intel	Cancel generation before its atomic save, app-only
mobile_log_sources	List connected Android devices and local Metro targets
mobile_logs_session	Start native and/or Metro log readers
mobile_read_logs	Read a log batch and source status
mobile_logs_keep_alive	Keep background collection alive without sending logs (app only)
mobile_logs_close	Stop a session's log readers
mobile_performance_sources	Read running apps and foreground identity on the selected iOS or Android device
mobile_cpu_session	Connect a native process and thread CPU plus memory monitor
mobile_read_cpu	Read live CPU and memory samples and connection status
mobile_cpu_close	Stop one CPU and memory monitor while leaving its app running
mobile_record_performance	Start a timed CPU and memory recording that saves automatically
mobile_read_performance_recording	Read original samples and a selected interval's summary
mobile_render_performance_recording	Show an interactive chart card in chat
mobile_compare_performance_recordings	Overlay 2–6 completed runs in an interactive comparison card
mobile_open_performance_recording	Open a saved run and selection in the workspace
mobile_finish_performance_recording	Stop and save a recording early
mobile_list_performance_recordings	Find recent saved and active runs
mobile_display_fps_session	Start device-wide Display FPS on Android 12+ or physical iOS 17.4+
mobile_read_display_fps	Read FPS intervals and connection status
mobile_read_performance_frames	Read Android jank/pacing statistics and page through exact display frames
mobile_display_fps_close	Stop FPS collection and release the tracing connection
mobile_appflow_start	Start a non-blocking structure build or screen sweep; Adaptive (the default) tries Runtime → Deep Link → verified replay without AI, while ai continues unresolved screens with interactive exploration. Legacy Fast, Auto and Thorough modes remain accepted.
mobile_appflow_status	Read current or persisted App Flow status
mobile_appflow_result	Read the current or persisted screen graph
mobile_appflow_events	Poll buffered progress and manual-recording events past a sequence cursor (app only)
mobile_appflow_pause	Pause a running sweep at a safe boundary (app only)
mobile_appflow_cancel	Cancel a run and any parked agent decision
mobile_appflow_launch_app	Launch the selected project without starting a sweep (app only)
mobile_appflow_resolve_prompt	Resolve a human-action gate after the user acts (app only)
mobile_appflow_projects	List the project folders App Flow can open (app only)
mobile_appflow_choose_project	Select a listed project or open the system folder picker (app only)
mobile_appflow_manual_start	Start recording a path to one screen (app only)
mobile_appflow_manual_act	Perform and record one manual path action (app only)
mobile_appflow_manual_stop	Save the active manual path recording (app only)
mobile_appflow_manual_cancel	Discard the active manual path recording (app only)
mobile_appflow_manual_replay	Replay a saved path to one screen (app only)
mobile_appflow_runtime_replay	Navigate to one screen through the running app's navigator, no recording needed (app only)
mobile_appflow_next_decision	Long-poll for the next active-agent sweep decision (model only)
mobile_appflow_submit_decision	Submit one role-valid JSON decision and wait briefly for the next (model only)
mobile_appflow_screenshot	Read an authorized App Flow PNG (app only)

The plugin servers for Codex, Claude and Cursor add the pinned agent-device runtime's commands as device_* tools with compact, validated schemas, including device_open, device_snapshot, device_screenshot, device_press, device_fill, device_type, device_scroll, device_wait, device_find, device_get, device_is, device_close, and debugging tools. The agent reads screens and sends input with them; device_screenshot also returns the saved image. Their annotations let hosts run reads, device input and app control without asking, as with the panel's tools, while installs, scripts, app-defined events, file outputs, the clipboard and development servers still ask (see Architecture). Their input schemas describe each command; logs and perf are left out for Mobile Dev's own tools. The agent-device skill explains the control loop and ships the runtime's own workflow guide.

TraceSift annotations

Analysis writes compact results and replaces the latest reference for the input profile. Remember writes/replaces the reference for a host view. Both declare readOnlyHint: false and destructiveHint: true because a previous association can be overwritten, even though the input profile is not edited.

Explore, Explore payload, and investigation handoff declare readOnlyHint: false and destructiveHint: false: they can start a local background service, and browser Explore also launches the explorer. The file picker shares those hints: it opens the host's own dialog and waits for a person, but writes nothing. The two restore tools only read local files and remain read-only. Reading the chosen project is read-only; choosing one declares destructiveHint: true, because it replaces the previously saved folder. All ten use openWorldHint: false; operations are confined to local files, the host's own dialogs, and the authenticated loopback TraceSift service.

The two pickers exist because the TraceSift tool runs in a panel frame, where a dropped file carries no path and the analyzer reads profiles off disk. Both are app-only: the model never calls them. The chosen profile path reaches mobile_analyze_profile from the panel, and the chosen project folder is the source tree an investigation lands in — a profile names frames, not files, so Investigate stays disabled until one is chosen, and the folder is named in the hand-off the panel sends to the chat.