API Reference

Built-in Tools

Reference for every built-in tool Mobile Agent exposes to models, grouped by capability.

Mobile Agent exposes a set of built-in tools that any connected model can call. Which tools are available is controlled per conversation and per agent through the approval system (see Tool Approvals).

Each tool has a stable key of the form camelCaseName. The full set:

Workspace

Operate on the app's private workspace directory.

Tool keyDescription
workspaceListFilesList files and directories in the workspace.
workspaceReadRead a file's contents.
workspaceWriteWrite content to a file (creates or overwrites).
workspaceCreateFileCreate a new empty file.
workspaceGrepSearch file contents with a regular expression.
workspaceGlobFind files matching a glob pattern.
workspaceEditApply a targeted edit to an existing file.

External folders

Operate on a mounted external folder session (Storage Access Framework).

Tool keyDescription
folderListDirectoryList entries in a mounted external folder.
folderReadRead a file from an external folder.
folderWriteWrite to a file in an external folder.
folderCreateFileCreate a file in an external folder.
folderCreateDirectoryCreate a directory in an external folder.
folderRenameEntryRename a file or directory.
folderMoveEntryMove a file or directory to a new location.
folderDeleteEntryDelete a file or directory.
folderGrepSearch external-folder file contents.
folderGlobFind external-folder files by glob.
folderEditApply a targeted edit in an external folder.

Files & downloads

Tool keyDescription
downloadFileDownload a URL into the workspace.

Conversation & orchestration

Tool keyDescription
todosRead and update the conversation's todo list (pending / in_progress / completed).
questionAsk you a structured question mid-run (freeform, choices, or multi-select).
skillActivate or manage a skill during the conversation.
schedulesCreate, update, or list scheduled jobs.
requestSecretPrompt you to configure an API key or token securely for a plugin or server.

Workflow notes

  • Tool approval — mutating tools (writes, deletes, downloads, terminal, schedules) pause the run in waiting_for_approval until approved. This gate cannot be disabled globally per tool beyond the per-conversation ask/auto mode.
  • Custom agents restrict which of these tools each agent may use — even if a tool is enabled globally, an agent with it disabled in its permissions cannot call it.
  • Skills alias mapping — some skills refer to tools by natural aliases; the app resolves common aliases (e.g. "file read", "grep") to the built-in keys above. This lets a skill written in plain language work without knowing exact keys.

Tool execution timeline

Every tool call is recorded in the message timeline with:

  • toolName and status (running / completed / failed)
  • inputSummary and outputSummary
  • error (when failed)
  • optional termux capture (command, output, taskId) when the tool ran a terminal command

Use the timeline to audit exactly what the agent did — the approval system controls the action, and the timeline records it.