replicant-mcp 1.2.1 → 1.3.0

package/dist/adapters/ui-automator.d.ts

@@ -66,6 +66,7 @@ export declare class UiAutomatorAdapter {
     tap(deviceId: string, x: number, y: number): Promise<void>;
     tapElement(deviceId: string, element: AccessibilityNode): Promise<void>;
     input(deviceId: string, text: string): Promise<void>;
+    scroll(deviceId: string, direction: "up" | "down" | "left" | "right", amount?: number): Promise<void>;
     screenshot(deviceId: string, options?: ScreenshotOptions): Promise<ScreenshotResult>;
     accessibilityCheck(deviceId: string): Promise<{
         hasAccessibleElements: boolean;

package/dist/adapters/ui-automator.js

@@ -82,6 +82,47 @@ export class UiAutomatorAdapter {
         const escaped = text.replace(/(['"\\$`])/g, "\\$1").replace(/ /g, "%s");
         await this.adb.shell(deviceId, `input text "${escaped}"`);
     }
+    async scroll(deviceId, direction, amount = 0.5) {
+        const screen = await this.getScreenMetadata(deviceId);
+        const { width, height } = screen;
+        // Calculate scroll distance based on amount (0-1 representing screen percentage)
+        const scrollDistance = Math.round((direction === "up" || direction === "down" ? height : width) * amount * 0.8);
+        // Center point of the screen
+        const centerX = Math.round(width / 2);
+        const centerY = Math.round(height / 2);
+        // Calculate start and end points based on direction
+        // Note: "scroll down" means content moves up, so we swipe up (finger moves from bottom to top)
+        let startX, startY, endX, endY;
+        switch (direction) {
+            case "down": // Scroll down = swipe up = finger moves up
+                startX = centerX;
+                startY = centerY + scrollDistance / 2;
+                endX = centerX;
+                endY = centerY - scrollDistance / 2;
+                break;
+            case "up": // Scroll up = swipe down = finger moves down
+                startX = centerX;
+                startY = centerY - scrollDistance / 2;
+                endX = centerX;
+                endY = centerY + scrollDistance / 2;
+                break;
+            case "right": // Scroll right = swipe left = finger moves left
+                startX = centerX + scrollDistance / 2;
+                startY = centerY;
+                endX = centerX - scrollDistance / 2;
+                endY = centerY;
+                break;
+            case "left": // Scroll left = swipe right = finger moves right
+                startX = centerX - scrollDistance / 2;
+                startY = centerY;
+                endX = centerX + scrollDistance / 2;
+                endY = centerY;
+                break;
+        }
+        // Duration in ms - longer for larger scrolls, minimum 100ms
+        const duration = Math.max(100, Math.round(scrollDistance / 2));
+        await this.adb.shell(deviceId, `input swipe ${startX} ${startY} ${endX} ${endY} ${duration}`);
+    }
     async screenshot(deviceId, options = {}) {
         const remotePath = "/sdcard/replicant-screenshot.png";
         const maxDimension = options.maxDimension ?? 1000;

package/dist/tools/ui.d.ts

@@ -7,6 +7,7 @@ export declare const uiInputSchema: z.ZodObject<{
         dump: "dump";
         tap: "tap";
         input: "input";
+        scroll: "scroll";
         screenshot: "screenshot";
         "accessibility-check": "accessibility-check";
         "visual-snapshot": "visual-snapshot";
@@ -29,6 +30,14 @@ export declare const uiInputSchema: z.ZodObject<{
     gridPosition: z.ZodOptional<z.ZodNumber>;
     maxDimension: z.ZodOptional<z.ZodNumber>;
     raw: z.ZodOptional<z.ZodBoolean>;
+    compact: z.ZodOptional<z.ZodBoolean>;
+    direction: z.ZodOptional<z.ZodEnum<{
+        up: "up";
+        down: "down";
+        left: "left";
+        right: "right";
+    }>>;
+    amount: z.ZodOptional<z.ZodNumber>;
 }, z.core.$strip>;
 export type UiInput = z.infer<typeof uiInputSchema>;
 export declare function handleUiTool(input: UiInput, context: ServerContext, uiConfig?: UiConfig): Promise<Record<string, unknown>>;
@@ -112,6 +121,21 @@ export declare const uiToolDefinition: {
                 type: string;
                 description: string;
             };
+            compact: {
+                type: string;
+                description: string;
+            };
+            direction: {
+                type: string;
+                enum: string[];
+                description: string;
+            };
+            amount: {
+                type: string;
+                minimum: number;
+                maximum: number;
+                description: string;
+            };
         };
         required: string[];
     };

package/dist/tools/ui.js

@@ -2,7 +2,7 @@ import { z } from "zod";
 import { CACHE_TTLS } from "../types/index.js";
 import { flattenTree } from "../parsers/ui-dump.js";
 export const uiInputSchema = z.object({
-    operation: z.enum(["dump", "find", "tap", "input", "screenshot", "accessibility-check", "visual-snapshot"]),
+    operation: z.enum(["dump", "find", "tap", "input", "scroll", "screenshot", "accessibility-check", "visual-snapshot"]),
     selector: z.object({
         resourceId: z.string().optional(),
         text: z.string().optional(),
@@ -21,6 +21,9 @@ export const uiInputSchema = z.object({
     gridPosition: z.number().min(1).max(5).optional(),
     maxDimension: z.number().optional(),
     raw: z.boolean().optional(),
+    compact: z.boolean().optional(),
+    direction: z.enum(["up", "down", "left", "right"]).optional(),
+    amount: z.number().min(0).max(1).optional(),
 });
 // Store last find results for elementIndex reference
 // Updated to support accessibility, OCR, and grid elements
@@ -103,7 +106,20 @@ export async function handleUiTool(input, context, uiConfig) {
             // Cache the tree
             const dumpId = context.cache.generateId("ui-dump");
             context.cache.set(dumpId, { tree, deviceId }, "ui-dump", CACHE_TTLS.UI_TREE);
-            // Create a simplified view
+            if (input.compact) {
+                // Compact mode: flat list of interactive elements only
+                const flat = flattenTree(tree);
+                const interactive = flat.filter((n) => n.clickable || n.focusable);
+                const elements = interactive.map((n) => ({
+                    text: n.text || n.contentDesc || undefined,
+                    type: n.className.split(".").pop(),
+                    x: n.centerX,
+                    y: n.centerY,
+                    resourceId: n.resourceId ? n.resourceId.split("/").pop() : undefined,
+                }));
+                return { dumpId, elements, count: elements.length, deviceId };
+            }
+            // Full mode: hierarchical tree with all details
             const simplifyNode = (node, depth = 0) => ({
                 className: node.className.split(".").pop(),
                 text: node.text || undefined,
@@ -318,10 +334,18 @@ export async function handleUiTool(input, context, uiConfig) {
             await context.ui.input(deviceId, input.text);
             return { input: input.text, deviceId };
         }
+        case "scroll": {
+            if (!input.direction) {
+                throw new Error("direction is required for scroll operation");
+            }
+            const amount = input.amount ?? 0.5; // Default to half-screen scroll
+            await context.ui.scroll(deviceId, input.direction, amount);
+            return { scrolled: { direction: input.direction, amount }, deviceId };
+        }
         case "screenshot": {
             const result = await context.ui.screenshot(deviceId, {
                 localPath: input.localPath,
-                inline: input.inline,
+                inline: input.inline ?? true,
                 maxDimension: input.maxDimension ?? config.maxImageDimension,
                 raw: input.raw,
             });
@@ -343,13 +367,13 @@ export async function handleUiTool(input, context, uiConfig) {
 }
 export const uiToolDefinition = {
     name: "ui",
-    description: "Interact with app UI via accessibility tree. Auto-selects device if only one connected. Operations: dump, find, tap, input, screenshot, accessibility-check, visual-snapshot.",
+    description: "Interact with app UI via accessibility tree. Auto-selects device if only one connected. Operations: dump, find, tap, input, scroll, screenshot, accessibility-check, visual-snapshot.",
     inputSchema: {
         type: "object",
         properties: {
             operation: {
                 type: "string",
-                enum: ["dump", "find", "tap", "input", "screenshot", "accessibility-check", "visual-snapshot"],
+                enum: ["dump", "find", "tap", "input", "scroll", "screenshot", "accessibility-check", "visual-snapshot"],
             },
             selector: {
                 type: "object",
@@ -367,7 +391,7 @@ export const uiToolDefinition = {
             elementIndex: { type: "number", description: "Element index from last find (for tap)" },
             text: { type: "string", description: "Text to input" },
             localPath: { type: "string", description: "Local path for screenshot (default: .replicant/screenshots/screenshot-{timestamp}.png)" },
-            inline: { type: "boolean", description: "Return base64 instead of file path (token-heavy, use sparingly)" },
+            inline: { type: "boolean", description: "Return base64 image data (default: true). Set to false to save to file instead." },
             debug: { type: "boolean", description: "Include source (accessibility/ocr) and confidence in response" },
             gridCell: { type: "number", minimum: 1, maximum: 24, description: "Grid cell number (1-24) for Tier 5 refinement" },
             gridPosition: { type: "number", minimum: 1, maximum: 5, description: "Position within cell (1=TL, 2=TR, 3=Center, 4=BL, 5=BR)" },
@@ -379,6 +403,21 @@ export const uiToolDefinition = {
                 type: "boolean",
                 description: "Skip scaling, return full device resolution. Warning: may exceed API limits.",
             },
+            compact: {
+                type: "boolean",
+                description: "For dump: return flat list of interactive elements with {text, type, x, y, resourceId} instead of full tree.",
+            },
+            direction: {
+                type: "string",
+                enum: ["up", "down", "left", "right"],
+                description: "Scroll direction (for scroll operation)",
+            },
+            amount: {
+                type: "number",
+                minimum: 0,
+                maximum: 1,
+                description: "Scroll amount as fraction of screen (0-1, default: 0.5)",
+            },
         },
         required: ["operation"],
     },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "replicant-mcp",
-  "version": "1.2.1",
+  "version": "1.3.0",
   "description": "Android MCP server for AI-assisted Android development",
   "type": "module",
   "main": "dist/index.js",
@@ -20,7 +20,10 @@
     "test:device": "tsx scripts/real-device-test.ts",
     "start": "npm run build && node dist/index.js",
     "validate": "npm run build && npm run test -- --run",
-    "prepublishOnly": "npm run build && npm test -- --run"
+    "prepublishOnly": "npm run build && npm test -- --run",
+    "release": "bash scripts/release.sh",
+    "release:minor": "bash scripts/release.sh minor",
+    "release:major": "bash scripts/release.sh major"
   },
   "keywords": [
     "mcp",