Android MCP Server

Instructions

Implementation Reference

• src/handlers.ts:528-556 (handler)

  Main tool handler that performs input validation and delegates to ADBWrapper.scrollInElement for execution.
  export async function uiautomatorScrollInElementHandler( adb: ADBWrapper, args: any ): Promise<{ content: Array<{ type: string; text: string }> }> { const { resourceId, direction, distance = 500, deviceSerial } = args as UIAutomatorScrollInElementArgs; if (!resourceId || typeof resourceId !== 'string') { throw new Error('Invalid resource ID: resourceId must be a non-empty string'); } if (!direction || !['up', 'down', 'left', 'right'].includes(direction)) { throw new Error('Invalid direction: must be one of "up", "down", "left", "right"'); } try { await adb.scrollInElement(resourceId, direction, distance, deviceSerial); return { content: [ { type: 'text', text: `Successfully scrolled ${direction} in element with resource-id: ${resourceId}\nDistance: ${distance}px`, }, ], }; } catch (error) { throw new Error(`UIAutomator scroll in element failed: ${error instanceof Error ? error.message : String(error)}`); } }
• src/adb-wrapper.ts:729-779 (helper)

  Core implementation: Dumps UI hierarchy XML, parses bounds of target element, computes relative swipe start/end points from element center based on direction, performs swipe gesture.
  async scrollInElement( resourceId: string, direction: 'up' | 'down' | 'left' | 'right', distance: number = 500, deviceSerial?: string ): Promise<void> { const device = await this.getTargetDevice(deviceSerial); const hierarchyFile = '/sdcard/window_dump.xml'; await this.exec(['shell', 'uiautomator', 'dump', hierarchyFile], device); const { stdout } = await this.exec(['shell', 'cat', hierarchyFile], device); await this.exec(['shell', 'rm', hierarchyFile], device); const boundsRegex = new RegExp(`resource-id="${resourceId}"[^>]*bounds="\\[(\\d+),(\\d+)\\]\\[(\\d+),(\\d+)\\]"`); const match = stdout.match(boundsRegex); if (match) { const x1 = parseInt(match[1], 10); const y1 = parseInt(match[2], 10); const x2 = parseInt(match[3], 10); const y2 = parseInt(match[4], 10); const centerX = Math.floor((x1 + x2) / 2); const centerY = Math.floor((y1 + y2) / 2); let startX = centerX, startY = centerY, endX = centerX, endY = centerY; switch (direction) { case 'up': startY = centerY + distance; endY = centerY - distance; break; case 'down': startY = centerY - distance; endY = centerY + distance; break; case 'left': startX = centerX + distance; endX = centerX - distance; break; case 'right': startX = centerX - distance; endX = centerX + distance; break; } await this.swipe(startX, startY, endX, endY, 300, device); } else { throw new Error(`Element with resource-id ${resourceId} not found in UI hierarchy`); } }
• src/index.ts:327-349 (schema)

  JSON schema defining input parameters for the tool, registered in listTools response.
  inputSchema: { type: 'object', properties: { resourceId: { type: 'string', description: 'Resource ID of the scrollable element', }, direction: { type: 'string', enum: ['up', 'down', 'left', 'right'], description: 'Direction to scroll', }, distance: { type: 'number', description: 'Distance to scroll in pixels (default: 500)', default: 500, }, deviceSerial: { type: 'string', description: 'Specific device serial number to target (optional)', }, }, required: ['resourceId', 'direction'],
• src/index.ts:492-493 (registration)

  Tool name dispatch in the switch statement handling CallToolRequestSchema.
  case 'android_uiautomator_scroll_in_element': return await uiautomatorScrollInElementHandler(this.adb, args);
• src/handlers.ts:81-86 (schema)

  TypeScript interface for input argument validation in the handler.
  interface UIAutomatorScrollInElementArgs { resourceId: string; direction: 'up' | 'down' | 'left' | 'right'; distance?: number; deviceSerial?: string; }