@nbakka/mcp-appium 2.0.5 → 2.0.8

package/lib/server.js

@@ -146,7 +146,6 @@ const createMcpServer = () => {
 tool("mobile_list_elements_on_screen", "List elements on screen and their coordinates, with display text or accessibility label. Do not cache this result.", {}, async ({}) => {
         requireRobot();
         const elements = await robot.getElementsOnScreen();
-	console.log(elements)
         const result = elements.map(element => {
             const out = {
                 type: element.type,
@@ -215,83 +214,73 @@ await new Promise(resolve => setTimeout(resolve, 5000));
         return `Current device orientation is ${orientation}`;
     });
 tool(
-  "mobile_create_session",
-  "create a mobile session once so that session id can be used in other tools where it is needed",
-  {},
-  async () => {
-    const capabilities = {
-      platformName: "Android",
-      "appium:udid": "emulator-5554",
-      "appium:automationName": "UiAutomator2",
-      "appium:noReset": true,
-      "appium:appPackage": "com.locon.housing",
-      "appium:appActivity": "com.locon.housing.presentation.MainActivity",
-    };
-
-    const payload = {
-      capabilities: {
-        firstMatch: [{}],
-        alwaysMatch: capabilities,
-      },
-    };
+  "tap_by_text",
+  "tap an element by passing the text visible on screen",
+  { text: "string" },
+  async ({ text }) => {
+    const { execSync } = require("child_process");
+    const { XMLParser } = require("fast-xml-parser");
 
-    const response = await fetch("http://localhost:4723/session", {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify(payload),
-    });
+    if (!text) throw new Error("Input text is required");
 
-    if (!response.ok) {
-      throw new Error(`Failed to create session: ${response.statusText}`);
+    // 1. Dump UI Automator XML to stdout
+    let dump;
+    try {
+      dump = execSync("adb shell uiautomator dump /dev/tty", { maxBuffer: 10 * 1024 * 1024 }).toString();
+    } catch (e) {
+      throw new Error("Failed to dump UI Automator XML via adb");
     }
 
-    const json = await response.json();
-    return `Session created with sessionId: ${json.sessionId}`;
-  }
-);
+    // 2. Parse XML
+    const parser = new XMLParser({
+      ignoreAttributes: false,
+      attributeNamePrefix: "",
+    });
+    const xmlObj = parser.parse(dump);
 
+    // 3. Recursive traversal to find matching element by text/content-desc/hint
+    function findElement(node) {
+      if (!node) return null;
 
-tool(
-  "mobile_click",
-  "Click an element identified by text using path",
-  {
-    sessionId: zod_1.z.string().describe("Appium session ID"),
-    text: zod_1.z.string().describe("Visible text of the element to click"),
-  },
-  async ({ sessionId, text }) => {
-    const xpath = `//*[@text="${text}"]`;
-    const clickUrl = `http://localhost:4723/session/${sessionId}/element`;
+      const matchText = node.text || node["content-desc"] || node.hint || "";
+      if (matchText === text) return node;
 
-    // Find element
-    const findResponse = await fetch(clickUrl, {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({ using: "xpath", value: xpath }),
-    });
+      if (node.node) {
+        if (Array.isArray(node.node)) {
+          for (const child of node.node) {
+            const found = findElement(child);
+            if (found) return found;
+          }
+        } else {
+          return findElement(node.node);
+        }
+      }
 
-    if (!findResponse.ok) {
-      throw new Error(`Failed to find element: ${findResponse.statusText}`);
+      return null;
     }
 
-    const findJson = await findResponse.json();
-    if (!findJson.value || !findJson.value.elementId) {
-      throw new Error(`Element with text "${text}" not found`);
-    }
+    const element = findElement(xmlObj.hierarchy.node);
+    if (!element) throw new Error(`Element with text "${text}" not found`);
 
-    const elementId = findJson.value.elementId;
+    // 4. Parse bounds: format "[left,top][right,bottom]"
+    const bounds = element.bounds;
+    if (!bounds) throw new Error("Element bounds not found");
 
-    // Click element
-    const clickElementUrl = `http://localhost:4723/session/${sessionId}/element/${elementId}/click`;
-    const clickResponse = await fetch(clickElementUrl, { method: "POST" });
+    const coords = bounds.match(/\d+/g).map(Number);
+    if (coords.length !== 4) throw new Error("Invalid bounds format");
 
-    if (!clickResponse.ok) {
-      throw new Error(`Failed to click element: ${clickResponse.statusText}`);
-    }
+    const [left, top, right, bottom] = coords;
+    const x = Math.floor((left + right) / 2);
+    const y = Math.floor((top + bottom) / 2);
 
-    // Optional wait after click
-    await new Promise((resolve) => setTimeout(resolve, 2000));
+    // 5. Tap via adb
+    try {
+      execSync(`adb shell input tap ${x} ${y}`);
+    } catch (e) {
+      throw new Error("Failed to perform tap via adb");
+    }
 
-    return `Clicked on element with text: "${text}" in session: ${sessionId}`;
+    return `Tapped element with text "${text}" at (${x},${y})`;
   }
 );
     // async check for latest agent version

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nbakka/mcp-appium",
-  "version": "2.0.5",
+  "version": "2.0.8",
   "description": "Appium MCP",
   "engines": {
     "node": ">=18"