@wdio/mcp 1.5.0 → 1.6.0

package/lib/server.js

@@ -7,13 +7,21 @@ import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js"
 // src/tools/browser.tool.ts
 import { remote } from "webdriverio";
 import { z } from "zod";
-var startBrowserToolArguments = {
-  headless: z.boolean().optional(),
-  windowWidth: z.number().min(400).max(3840).optional(),
-  windowHeight: z.number().min(400).max(2160).optional()
+var startBrowserToolDefinition = {
+  name: "start_browser",
+  description: "starts a browser session and sets it to the current state",
+  inputSchema: {
+    headless: z.boolean().optional(),
+    windowWidth: z.number().min(400).max(3840).optional(),
+    windowHeight: z.number().min(400).max(2160).optional()
+  }
 };
-var closeSessionToolArguments = {
-  detach: z.boolean().optional().describe("If true, disconnect from session without terminating it (preserves app state). Default: false")
+var closeSessionToolDefinition = {
+  name: "close_session",
+  description: "closes or detaches from the current browser or app session",
+  inputSchema: {
+    detach: z.boolean().optional().describe("If true, disconnect from session without terminating it (preserves app state). Default: false")
+  }
 };
 var state = {
   browsers: /* @__PURE__ */ new Map(),
@@ -95,8 +103,12 @@ var closeSessionTool = async (args = {}) => {
 
 // src/tools/navigate.tool.ts
 import { z as z2 } from "zod";
-var navigateToolArguments = {
-  url: z2.string().nonempty("URL must be provided")
+var navigateToolDefinition = {
+  name: "navigate",
+  description: "navigates to a URL",
+  inputSchema: {
+    url: z2.string().min(1).describe("The URL to navigate to")
+  }
 };
 var navigateTool = async ({ url }) => {
   try {
@@ -115,10 +127,23 @@ var navigateTool = async ({ url }) => {
 // src/tools/click.tool.ts
 import { z as z3 } from "zod";
 var defaultTimeout = 3e3;
-var clickToolArguments = {
-  selector: z3.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']" or "button=Exact text with spaces" or "a*=Link containing text")`),
-  scrollToView: z3.boolean().optional().describe("Whether to scroll the element into view before clicking").default(true),
-  timeout: z3.number().optional().describe("Maximum time to wait for element in milliseconds")
+var clickToolDefinition = {
+  name: "click_element",
+  description: "clicks an element",
+  inputSchema: {
+    selector: z3.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']" or "button=Exact text with spaces" or "a*=Link containing text")`),
+    scrollToView: z3.boolean().optional().describe("Whether to scroll the element into view before clicking").default(true),
+    timeout: z3.number().optional().describe("Maximum time to wait for element in milliseconds")
+  }
+};
+var clickViaTextToolDefinition = {
+  name: "click_via_text",
+  description: "clicks an element",
+  inputSchema: {
+    selector: z3.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']" or "button=Exact text with spaces" or "a*=Link containing text")`),
+    scrollToView: z3.boolean().optional().describe("Whether to scroll the element into view before clicking").default(true),
+    timeout: z3.number().optional().describe("Maximum time to wait for element in milliseconds")
+  }
 };
 var clickAction = async (selector, timeout, scrollToView = true) => {
   try {
@@ -143,11 +168,15 @@ var clickToolViaText = async ({ text, scrollToView, timeout = defaultTimeout })
 // src/tools/set-value.tool.ts
 import { z as z4 } from "zod";
 var defaultTimeout2 = 3e3;
-var setValueToolArguments = {
-  selector: z4.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']")`),
-  value: z4.string().describe("Text to enter into the element"),
-  scrollToView: z4.boolean().optional().describe("Whether to scroll the element into view before typing").default(true),
-  timeout: z4.number().optional().describe("Maximum time to wait for element in milliseconds")
+var setValueToolDefinition = {
+  name: "set_value",
+  description: "set value to an element, aka typing",
+  inputSchema: {
+    selector: z4.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']")`),
+    value: z4.string().describe("Text to enter into the element"),
+    scrollToView: z4.boolean().optional().describe("Whether to scroll the element into view before typing").default(true),
+    timeout: z4.number().optional().describe("Maximum time to wait for element in milliseconds")
+  }
 };
 var setValueTool = async ({ selector, value, scrollToView = true, timeout = defaultTimeout2 }) => {
   try {
@@ -171,9 +200,13 @@ var setValueTool = async ({ selector, value, scrollToView = true, timeout = defa
 // src/tools/find-element.tool.ts
 import { z as z5 } from "zod";
 var defaultTimeout3 = 3e3;
-var findElementToolArguments = {
-  selector: z5.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']")`),
-  timeout: z5.number().optional().describe("Maximum time to wait for element in milliseconds")
+var findElementToolDefinition = {
+  name: "find_element",
+  description: "finds an element",
+  inputSchema: {
+    selector: z5.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']")`),
+    timeout: z5.number().optional().describe("Maximum time to wait for element in milliseconds")
+  }
 };
 var findElementTool = async ({ selector, timeout = defaultTimeout3 }) => {
   try {
@@ -192,9 +225,13 @@ var findElementTool = async ({ selector, timeout = defaultTimeout3 }) => {
 // src/tools/get-element-text.tool.ts
 import { z as z6 } from "zod";
 var defaultTimeout4 = 3e3;
-var getElementTextToolArguments = {
-  selector: z6.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']")`),
-  timeout: z6.number().optional().describe("Maximum time to wait for element in milliseconds")
+var getElementTextToolDefinition = {
+  name: "get_element_text",
+  description: "gets the text content of an element",
+  inputSchema: {
+    selector: z6.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']")`),
+    timeout: z6.number().optional().describe("Maximum time to wait for element in milliseconds")
+  }
 };
 var getElementTextTool = async ({ selector, timeout = defaultTimeout4 }) => {
   try {
@@ -214,9 +251,13 @@ var getElementTextTool = async ({ selector, timeout = defaultTimeout4 }) => {
 // src/tools/is-displayed.tool.ts
 import { z as z7 } from "zod";
 var defaultTimeout5 = 3e3;
-var isDisplayedToolArguments = {
-  selector: z7.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']")`),
-  timeout: z7.number().optional().describe("Maximum time to wait for element in milliseconds")
+var isDisplayedToolDefinition = {
+  name: "is_displayed",
+  description: "checks if an element is displayed",
+  inputSchema: {
+    selector: z7.string().describe(`Value for the selector, in the form of css selector or xpath ("button.my-class" or "//button[@class='my-class']")`),
+    timeout: z7.number().optional().describe("Maximum time to wait for element in milliseconds")
+  }
 };
 var isDisplayedTool = async ({ selector, timeout = defaultTimeout5 }) => {
   try {
@@ -238,8 +279,12 @@ var isDisplayedTool = async ({ selector, timeout = defaultTimeout5 }) => {
 
 // src/tools/scroll-down.tool.ts
 import { z as z8 } from "zod";
-var scrollDownToolArguments = {
-  pixels: z8.number().optional().default(500)
+var scrollDownToolDefinition = {
+  name: "scroll_down",
+  description: "scrolls the page down by specified pixels",
+  inputSchema: {
+    pixels: z8.number().optional().default(500)
+  }
 };
 var scrollDownTool = async ({ pixels = 500 }) => {
   try {
@@ -259,8 +304,12 @@ var scrollDownTool = async ({ pixels = 500 }) => {
 
 // src/tools/scroll-up.tool.ts
 import { z as z9 } from "zod";
-var scrollUpToolArguments = {
-  pixels: z9.number().optional().default(500)
+var scrollUpToolDefinition = {
+  name: "scroll_up",
+  description: "scrolls the page up by specified pixels",
+  inputSchema: {
+    pixels: z9.number().optional().default(500)
+  }
 };
 var scrollUpTool = async ({ pixels = 500 }) => {
   try {
@@ -1049,24 +1098,66 @@ function generateAllElementLocators(sourceXML, options) {
 }
 
 // src/utils/mobile-elements.ts
-function toMobileElementInfo(element) {
-  const locatorValues = Object.values(element.locators);
-  return {
-    selector: locatorValues[0] || "",
-    alternativeSelectors: locatorValues.length > 1 ? locatorValues.slice(1, 3) : void 0,
-    locators: element.locators,
-    isInViewport: element.isInViewport,
-    isEnabled: element.enabled,
-    bounds: element.bounds,
-    tagName: element.tagName || void 0,
-    text: element.text || void 0,
-    resourceId: element.resourceId || void 0,
-    contentDesc: element.contentDesc || void 0,
-    accessibilityId: element.accessibilityId || void 0,
-    label: element.label || void 0,
-    value: element.value || void 0,
-    className: element.className || void 0
+var LOCATOR_PRIORITY = [
+  "accessibility-id",
+  // Most stable, cross-platform
+  "id",
+  // Android resource-id
+  "text",
+  // Text-based (can be fragile but readable)
+  "predicate-string",
+  // iOS predicate
+  "class-chain",
+  // iOS class chain
+  "uiautomator",
+  // Android UiAutomator compound
+  "xpath"
+  // XPath (last resort, brittle)
+  // 'class-name' intentionally excluded - too generic
+];
+function selectBestLocators(locators) {
+  const selected = [];
+  for (const strategy of LOCATOR_PRIORITY) {
+    if (locators[strategy]) {
+      selected.push(locators[strategy]);
+      break;
+    }
+  }
+  for (const strategy of LOCATOR_PRIORITY) {
+    if (locators[strategy] && !selected.includes(locators[strategy])) {
+      selected.push(locators[strategy]);
+      break;
+    }
+  }
+  return selected;
+}
+function toMobileElementInfo(element, includeBounds) {
+  const selectedLocators = selectBestLocators(element.locators);
+  const info = {
+    selector: selectedLocators[0] || "",
+    tagName: element.tagName,
+    isInViewport: element.isInViewport
   };
+  if (element.text) {
+    info.text = element.text;
+  }
+  if (element.resourceId) {
+    info.resourceId = element.resourceId;
+  }
+  const accessId = element.accessibilityId || element.contentDesc;
+  if (accessId) {
+    info.accessibilityId = accessId;
+  }
+  if (!element.enabled) {
+    info.isEnabled = false;
+  }
+  if (selectedLocators.length > 1) {
+    info.alternativeSelectors = selectedLocators.slice(1);
+  }
+  if (includeBounds) {
+    info.bounds = element.bounds;
+  }
+  return info;
 }
 async function getViewportSize(browser) {
   try {
@@ -1077,7 +1168,7 @@ async function getViewportSize(browser) {
   }
 }
 async function getMobileVisibleElements(browser, platform, options = {}) {
-  const { includeContainers = false, filterOptions } = options;
+  const { includeContainers = false, includeBounds = false, filterOptions } = options;
   const viewportSize = await getViewportSize(browser);
   const pageSource = await browser.getPageSource();
   const filters = {
@@ -1089,7 +1180,7 @@ async function getMobileVisibleElements(browser, platform, options = {}) {
     viewportSize,
     filters
   });
-  return elements.map(toMobileElementInfo);
+  return elements.map((el) => toMobileElementInfo(el, includeBounds));
 }
 
 // src/tools/get-visible-elements.tool.ts
@@ -1107,19 +1198,25 @@ function stripUndefinedFromArray(arr) {
 }
 
 // src/tools/get-visible-elements.tool.ts
-var getVisibleElementsToolArguments = {
-  inViewportOnly: z10.boolean().optional().describe(
-    "Only return elements within the visible viewport. Default: true. Set to false to get ALL elements on the page."
-  ),
-  includeContainers: z10.boolean().optional().describe(
-    "Include layout containers (ViewGroup, FrameLayout, ScrollView, etc). Default: false. Set to true to see all elements including layouts."
-  ),
-  elementType: z10.enum(["interactable", "visual", "all"]).optional().describe(
-    'Type of elements to return: "interactable" (default) for buttons/links/inputs, "visual" for images/SVGs, "all" for both.'
-  ),
-  limit: z10.number().optional().describe(
-    "Maximum number of elements to return. Default: 0 (unlimited). Set a limit for pages with many elements."
-  )
+var getVisibleElementsToolDefinition = {
+  name: "get_visible_elements",
+  description: 'get a list of visible (in viewport & displayed) interactable elements on the page (buttons, links, inputs). Use elementType="visual" for images/SVGs. Must prefer this to take_screenshot for interactions',
+  inputSchema: {
+    inViewportOnly: z10.boolean().optional().describe(
+      "Only return elements within the visible viewport. Default: true. Set to false to get ALL elements on the page."
+    ),
+    includeContainers: z10.boolean().optional().describe(
+      "Include layout containers (ViewGroup, FrameLayout, ScrollView, etc). Default: false. Set to true to see all elements including layouts."
+    ),
+    includeBounds: z10.boolean().optional().describe(
+      "Include element bounds/coordinates (x, y, width, height). Default: false. Set to true for coordinate-based interactions or layout debugging."
+    ),
+    elementType: z10.enum(["interactable", "visual", "all"]).optional().describe(
+      'Type of elements to return: "interactable" (default) for buttons/links/inputs, "visual" for images/SVGs, "all" for both.'
+    ),
+    limit: z10.number().optional().describe("Maximum number of elements to return. Default: 0 (unlimited)."),
+    offset: z10.number().optional().describe("Number of elements to skip (for pagination). Default: 0.")
+  }
 };
 var getVisibleElementsTool = async (args) => {
   try {
@@ -1127,34 +1224,37 @@ var getVisibleElementsTool = async (args) => {
     const {
       inViewportOnly = true,
       includeContainers = false,
+      includeBounds = false,
       elementType = "interactable",
-      limit = 0
+      limit = 0,
+      offset = 0
     } = args || {};
+    let elements;
     if (browser.isAndroid || browser.isIOS) {
       const platform = browser.isAndroid ? "android" : "ios";
-      let elements2 = await getMobileVisibleElements(browser, platform, {
-        includeContainers
-      });
-      if (inViewportOnly) {
-        elements2 = elements2.filter((el) => el.isInViewport);
-      }
-      if (limit > 0 && elements2.length > limit) {
-        elements2 = elements2.slice(0, limit);
-      }
-      return {
-        content: [{ type: "text", text: encode(elements2) }]
-      };
+      elements = await getMobileVisibleElements(browser, platform, { includeContainers, includeBounds });
+    } else {
+      const raw = await browser.execute(get_interactable_elements_default, elementType);
+      elements = stripUndefinedFromArray(raw);
     }
-    let elements = await browser.execute(get_interactable_elements_default, elementType);
     if (inViewportOnly) {
       elements = elements.filter((el) => el.isInViewport !== false);
     }
-    if (limit > 0 && elements.length > limit) {
+    const total = elements.length;
+    if (offset > 0) {
+      elements = elements.slice(offset);
+    }
+    if (limit > 0) {
       elements = elements.slice(0, limit);
     }
-    const cleanedElements = stripUndefinedFromArray(elements);
+    const result = {
+      total,
+      showing: elements.length,
+      hasMore: offset + elements.length < total,
+      elements
+    };
     return {
-      content: [{ type: "text", text: encode(cleanedElements) }]
+      content: [{ type: "text", text: encode(result) }]
     };
   } catch (e) {
     return {
@@ -1165,8 +1265,12 @@ var getVisibleElementsTool = async (args) => {
 
 // src/tools/take-screenshot.tool.ts
 import { z as z11 } from "zod";
-var takeScreenshotToolArguments = {
-  outputPath: z11.string().optional().describe("Optional path where to save the screenshot. If not provided, returns base64 data.")
+var takeScreenshotToolDefinition = {
+  name: "take_screenshot",
+  description: "captures a screenshot of the current page",
+  inputSchema: {
+    outputPath: z11.string().optional().describe("Optional path where to save the screenshot. If not provided, returns base64 data.")
+  }
 };
 var takeScreenshotTool = async ({ outputPath }) => {
   try {
@@ -1194,8 +1298,12 @@ var takeScreenshotTool = async ({ outputPath }) => {
 
 // src/tools/cookies.tool.ts
 import { z as z12 } from "zod";
-var getCookiesToolArguments = {
-  name: z12.string().optional().describe("Optional cookie name to retrieve a specific cookie. If not provided, returns all cookies")
+var getCookiesToolDefinition = {
+  name: "get_cookies",
+  description: "gets all cookies or a specific cookie by name",
+  inputSchema: {
+    name: z12.string().optional().describe("Optional cookie name to retrieve a specific cookie. If not provided, returns all cookies")
+  }
 };
 var getCookiesTool = async ({ name }) => {
   try {
@@ -1226,15 +1334,19 @@ var getCookiesTool = async ({ name }) => {
     };
   }
 };
-var setCookieToolArguments = {
-  name: z12.string().describe("Cookie name"),
-  value: z12.string().describe("Cookie value"),
-  domain: z12.string().optional().describe("Cookie domain (defaults to current domain)"),
-  path: z12.string().optional().describe('Cookie path (defaults to "/")'),
-  expires: z12.number().optional().describe("Expiry date as Unix timestamp in seconds"),
-  httpOnly: z12.boolean().optional().describe("HttpOnly flag"),
-  secure: z12.boolean().optional().describe("Secure flag"),
-  sameSite: z12.enum(["Strict", "Lax", "None"]).optional().describe("SameSite attribute")
+var setCookieToolDefinition = {
+  name: "set_cookie",
+  description: "sets a cookie with specified name, value, and optional attributes",
+  inputSchema: {
+    name: z12.string().describe("Cookie name"),
+    value: z12.string().describe("Cookie value"),
+    domain: z12.string().optional().describe("Cookie domain (defaults to current domain)"),
+    path: z12.string().optional().describe('Cookie path (defaults to "/")'),
+    expires: z12.number().optional().describe("Expiry date as Unix timestamp in seconds"),
+    httpOnly: z12.boolean().optional().describe("HttpOnly flag"),
+    secure: z12.boolean().optional().describe("Secure flag"),
+    sameSite: z12.enum(["Strict", "Lax", "None"]).optional().describe("SameSite attribute")
+  }
 };
 var setCookieTool = async ({
   name,
@@ -1268,8 +1380,12 @@ var setCookieTool = async ({
     };
   }
 };
-var deleteCookiesToolArguments = {
-  name: z12.string().optional().describe("Optional cookie name to delete a specific cookie. If not provided, deletes all cookies")
+var deleteCookiesToolDefinition = {
+  name: "delete_cookies",
+  description: "deletes all cookies or a specific cookie by name",
+  inputSchema: {
+    name: z12.string().optional().describe("Optional cookie name to delete a specific cookie. If not provided, deletes all cookies")
+  }
 };
 var deleteCookiesTool = async ({ name }) => {
   try {
@@ -1294,10 +1410,15 @@ var deleteCookiesTool = async ({ name }) => {
 // src/tools/get-accessibility-tree.tool.ts
 import { encode as encode2 } from "@toon-format/toon";
 import { z as z13 } from "zod";
-var getAccessibilityToolArguments = {
-  limit: z13.number().optional().describe("Maximum number of nodes to return. Default: 100. Use 0 for unlimited."),
-  roles: z13.array(z13.string()).optional().describe('Filter to specific roles (e.g., ["button", "link", "textbox"]). Default: all roles.'),
-  namedOnly: z13.boolean().optional().describe("Only return nodes with a name/label. Default: true. Filters out anonymous containers.")
+var getAccessibilityToolDefinition = {
+  name: "get_accessibility",
+  description: "gets accessibility tree snapshot with semantic information about page elements (roles, names, states). Browser-only - use when get_visible_elements does not return expected elements.",
+  inputSchema: {
+    limit: z13.number().optional().describe("Maximum number of nodes to return. Default: 100. Use 0 for unlimited."),
+    offset: z13.number().optional().describe("Number of nodes to skip (for pagination). Default: 0."),
+    roles: z13.array(z13.string()).optional().describe('Filter to specific roles (e.g., ["button", "link", "textbox"]). Default: all roles.'),
+    namedOnly: z13.boolean().optional().describe("Only return nodes with a name/label. Default: true. Filters out anonymous containers.")
+  }
 };
 function flattenAccessibilityTree(node, result = []) {
   if (!node) return result;
@@ -1348,7 +1469,7 @@ var getAccessibilityTreeTool = async (args) => {
         }]
       };
     }
-    const { limit = 100, roles, namedOnly = true } = args || {};
+    const { limit = 100, offset = 0, roles, namedOnly = true } = args || {};
     const puppeteer = await browser.getPuppeteer();
     const pages = await puppeteer.pages();
     if (pages.length === 0) {
@@ -1374,14 +1495,21 @@ var getAccessibilityTreeTool = async (args) => {
       const roleSet = new Set(roles.map((r) => r.toLowerCase()));
       nodes = nodes.filter((n) => n.role && roleSet.has(n.role.toLowerCase()));
     }
-    if (limit > 0 && nodes.length > limit) {
+    const total = nodes.length;
+    if (offset > 0) {
+      nodes = nodes.slice(offset);
+    }
+    if (limit > 0) {
       nodes = nodes.slice(0, limit);
     }
+    const result = {
+      total,
+      showing: nodes.length,
+      hasMore: offset + nodes.length < total,
+      nodes
+    };
     return {
-      content: [{
-        type: "text",
-        text: encode2(nodes)
-      }]
+      content: [{ type: "text", text: encode2(result) }]
     };
   } catch (e) {
     return {
@@ -1472,22 +1600,26 @@ function buildAndroidCapabilities(appPath, options) {
 }
 
 // src/tools/app-session.tool.ts
-var startAppToolArguments = {
-  platform: z14.enum(["iOS", "Android"]).describe("Mobile platform"),
-  appPath: z14.string().optional().describe("Path to the app file (.app/.apk/.ipa). Required unless noReset=true (connecting to already-running app)"),
-  deviceName: z14.string().describe("Device/emulator/simulator name"),
-  platformVersion: z14.string().optional().describe('OS version (e.g., "17.0", "14")'),
-  automationName: z14.enum(["XCUITest", "UiAutomator2", "Espresso"]).optional().describe("Automation driver name"),
-  appiumHost: z14.string().optional().describe("Appium server hostname (overrides APPIUM_URL env var)"),
-  appiumPort: z14.number().optional().describe("Appium server port (overrides APPIUM_URL_PORT env var)"),
-  appiumPath: z14.string().optional().describe("Appium server path (overrides APPIUM_PATH env var)"),
-  autoGrantPermissions: z14.boolean().optional().describe("Auto-grant app permissions (default: true)"),
-  autoAcceptAlerts: z14.boolean().optional().describe("Auto-accept alerts (default: true)"),
-  autoDismissAlerts: z14.boolean().optional().describe('Auto-dismiss alerts (default: false, will override "autoAcceptAlerts" to undefined if set)'),
-  appWaitActivity: z14.string().optional().describe("Activity to wait for on launch (Android only)"),
-  udid: z14.string().optional().describe('Unique Device Identifier for iOS real device testing (e.g., "00008030-001234567890002E")'),
-  noReset: z14.boolean().optional().describe("Do not reset app state before session (preserves app data). Default: false"),
-  fullReset: z14.boolean().optional().describe("Uninstall app before/after session. Default: true. Set to false with noReset=true to preserve app state completely")
+var startAppToolDefinition = {
+  name: "start_app_session",
+  description: "starts a mobile app session (iOS/Android) via Appium",
+  inputSchema: {
+    platform: z14.enum(["iOS", "Android"]).describe("Mobile platform"),
+    appPath: z14.string().optional().describe("Path to the app file (.app/.apk/.ipa). Required unless noReset=true (connecting to already-running app)"),
+    deviceName: z14.string().describe("Device/emulator/simulator name"),
+    platformVersion: z14.string().optional().describe('OS version (e.g., "17.0", "14")'),
+    automationName: z14.enum(["XCUITest", "UiAutomator2", "Espresso"]).optional().describe("Automation driver name"),
+    appiumHost: z14.string().optional().describe("Appium server hostname (overrides APPIUM_URL env var)"),
+    appiumPort: z14.number().optional().describe("Appium server port (overrides APPIUM_URL_PORT env var)"),
+    appiumPath: z14.string().optional().describe("Appium server path (overrides APPIUM_PATH env var)"),
+    autoGrantPermissions: z14.boolean().optional().describe("Auto-grant app permissions (default: true)"),
+    autoAcceptAlerts: z14.boolean().optional().describe("Auto-accept alerts (default: true)"),
+    autoDismissAlerts: z14.boolean().optional().describe('Auto-dismiss alerts (default: false, will override "autoAcceptAlerts" to undefined if set)'),
+    appWaitActivity: z14.string().optional().describe("Activity to wait for on launch (Android only)"),
+    udid: z14.string().optional().describe('Unique Device Identifier for iOS real device testing (e.g., "00008030-001234567890002E")'),
+    noReset: z14.boolean().optional().describe("Do not reset app state before session (preserves app data). Default: false"),
+    fullReset: z14.boolean().optional().describe("Uninstall app before/after session. Default: true. Set to false with noReset=true to preserve app state completely")
+  }
 };
 var getState = () => {
   const sharedState = getBrowser.__state;
@@ -1588,10 +1720,14 @@ Appium Server: ${serverConfig.hostname}:${serverConfig.port}${serverConfig.path}
 
 // src/tools/gestures.tool.ts
 import { z as z15 } from "zod";
-var tapElementToolArguments = {
-  selector: z15.string().optional().describe("Element selector (CSS, XPath, accessibility ID, or UiAutomator)"),
-  x: z15.number().optional().describe("X coordinate for tap (if no selector provided)"),
-  y: z15.number().optional().describe("Y coordinate for tap (if no selector provided)")
+var tapElementToolDefinition = {
+  name: "tap_element",
+  description: "taps an element by selector or coordinates (mobile)",
+  inputSchema: {
+    selector: z15.string().optional().describe("Element selector (CSS, XPath, accessibility ID, or UiAutomator)"),
+    x: z15.number().optional().describe("X coordinate for tap (if no selector provided)"),
+    y: z15.number().optional().describe("Y coordinate for tap (if no selector provided)")
+  }
 };
 var tapElementTool = async (args) => {
   try {
@@ -1622,12 +1758,16 @@ var tapElementTool = async (args) => {
     };
   }
 };
-var swipeToolArguments = {
-  direction: z15.enum(["up", "down", "left", "right"]).describe("Swipe direction"),
-  duration: z15.number().min(100).max(5e3).optional().describe("Swipe duration in milliseconds (default: 500)"),
-  startX: z15.number().optional().describe("Start X coordinate (optional, uses screen center)"),
-  startY: z15.number().optional().describe("Start Y coordinate (optional, uses screen center)"),
-  distance: z15.number().optional().describe("Swipe distance in pixels (optional, uses percentage of screen)")
+var swipeToolDefinition = {
+  name: "swipe",
+  description: "performs a swipe gesture in specified direction (mobile)",
+  inputSchema: {
+    direction: z15.enum(["up", "down", "left", "right"]).describe("Swipe direction"),
+    duration: z15.number().min(100).max(5e3).optional().describe("Swipe duration in milliseconds (default: 500)"),
+    startX: z15.number().optional().describe("Start X coordinate (optional, uses screen center)"),
+    startY: z15.number().optional().describe("Start Y coordinate (optional, uses screen center)"),
+    distance: z15.number().optional().describe("Swipe distance in pixels (optional, uses percentage of screen)")
+  }
 };
 var swipeTool = async (args) => {
   try {
@@ -1675,11 +1815,15 @@ var swipeTool = async (args) => {
     };
   }
 };
-var longPressToolArguments = {
-  selector: z15.string().optional().describe("Element selector (CSS, XPath, accessibility ID, or UiAutomator)"),
-  x: z15.number().optional().describe("X coordinate for long press (if no selector provided)"),
-  y: z15.number().optional().describe("Y coordinate for long press (if no selector provided)"),
-  duration: z15.number().min(500).max(1e4).optional().describe("Long press duration in milliseconds (default: 1000)")
+var longPressToolDefinition = {
+  name: "long_press",
+  description: "performs a long press on element or coordinates (mobile)",
+  inputSchema: {
+    selector: z15.string().optional().describe("Element selector (CSS, XPath, accessibility ID, or UiAutomator)"),
+    x: z15.number().optional().describe("X coordinate for long press (if no selector provided)"),
+    y: z15.number().optional().describe("Y coordinate for long press (if no selector provided)"),
+    duration: z15.number().min(500).max(1e4).optional().describe("Long press duration in milliseconds (default: 1000)")
+  }
 };
 var longPressTool = async (args) => {
   try {
@@ -1714,14 +1858,18 @@ var longPressTool = async (args) => {
     };
   }
 };
-var dragAndDropToolArguments = {
-  fromSelector: z15.string().optional().describe("Source element selector"),
-  fromX: z15.number().optional().describe("Source X coordinate"),
-  fromY: z15.number().optional().describe("Source Y coordinate"),
-  toSelector: z15.string().optional().describe("Target element selector"),
-  toX: z15.number().optional().describe("Target X coordinate"),
-  toY: z15.number().optional().describe("Target Y coordinate"),
-  duration: z15.number().min(100).max(5e3).optional().describe("Drag duration in milliseconds (default: 500)")
+var dragAndDropToolDefinition = {
+  name: "drag_and_drop",
+  description: "drags from one location to another (mobile)",
+  inputSchema: {
+    fromSelector: z15.string().optional().describe("Source element selector"),
+    fromX: z15.number().optional().describe("Source X coordinate"),
+    fromY: z15.number().optional().describe("Source Y coordinate"),
+    toSelector: z15.string().optional().describe("Target element selector"),
+    toX: z15.number().optional().describe("Target X coordinate"),
+    toY: z15.number().optional().describe("Target Y coordinate"),
+    duration: z15.number().min(100).max(5e3).optional().describe("Drag duration in milliseconds (default: 500)")
+  }
 };
 var dragAndDropTool = async (args) => {
   try {
@@ -1782,8 +1930,12 @@ var dragAndDropTool = async (args) => {
 
 // src/tools/app-actions.tool.ts
 import { z as z16 } from "zod";
-var getAppStateToolArguments = {
-  bundleId: z16.string().describe("App bundle ID (e.g., com.example.app)")
+var getAppStateToolDefinition = {
+  name: "get_app_state",
+  description: "gets the state of an app (not installed, not running, background, foreground)",
+  inputSchema: {
+    bundleId: z16.string().describe("App bundle ID (e.g., com.example.app)")
+  }
 };
 var getAppStateTool = async (args) => {
   try {
@@ -1812,8 +1964,12 @@ var getAppStateTool = async (args) => {
     };
   }
 };
-var activateAppToolArguments = {
-  bundleId: z16.string().describe("App bundle ID to activate (e.g., com.example.app)")
+var activateAppToolDefinition = {
+  name: "activate_app",
+  description: "activates/brings an app to foreground",
+  inputSchema: {
+    bundleId: z16.string().describe("App bundle ID to activate (e.g., com.example.app)")
+  }
 };
 var activateAppTool = async (args) => {
   try {
@@ -1830,8 +1986,12 @@ var activateAppTool = async (args) => {
     };
   }
 };
-var terminateAppToolArguments = {
-  bundleId: z16.string().describe("App bundle ID to terminate (e.g., com.example.app)")
+var terminateAppToolDefinition = {
+  name: "terminate_app",
+  description: "terminates a running app",
+  inputSchema: {
+    bundleId: z16.string().describe("App bundle ID to terminate (e.g., com.example.app)")
+  }
 };
 var terminateAppTool = async (args) => {
   try {
@@ -1851,6 +2011,25 @@ var terminateAppTool = async (args) => {
 
 // src/tools/context.tool.ts
 import { z as z17 } from "zod";
+var getContextsToolDefinition = {
+  name: "get_contexts",
+  description: "lists available contexts (NATIVE_APP, WEBVIEW)",
+  inputSchema: {}
+};
+var getCurrentContextToolDefinition = {
+  name: "get_current_context",
+  description: "shows the currently active context",
+  inputSchema: {}
+};
+var switchContextToolDefinition = {
+  name: "switch_context",
+  description: "switches between native and webview contexts",
+  inputSchema: {
+    context: z17.string().describe(
+      'Context name to switch to (e.g., "NATIVE_APP", "WEBVIEW_com.example.app", or use index from get_contexts)'
+    )
+  }
+};
 var getContextsTool = async () => {
   try {
     const browser = getBrowser();
@@ -1883,11 +2062,6 @@ var getCurrentContextTool = async () => {
     };
   }
 };
-var switchContextToolArguments = {
-  context: z17.string().describe(
-    'Context name to switch to (e.g., "NATIVE_APP", "WEBVIEW_com.example.app", or use index from get_contexts)'
-  )
-};
 var switchContextTool = async (args) => {
   try {
     const browser = getBrowser();
@@ -1922,6 +2096,86 @@ var switchContextTool = async (args) => {
 
 // src/tools/device.tool.ts
 import { z as z18 } from "zod";
+var getDeviceInfoToolDefinition = {
+  name: "get_device_info",
+  description: "gets device information (platform, version, screen size)",
+  inputSchema: {}
+};
+var getOrientationToolDefinition = {
+  name: "get_orientation",
+  description: "gets current device orientation",
+  inputSchema: {}
+};
+var lockDeviceToolDefinition = {
+  name: "lock_device",
+  description: "locks the device screen",
+  inputSchema: {}
+};
+var unlockDeviceToolDefinition = {
+  name: "unlock_device",
+  description: "unlocks the device screen",
+  inputSchema: {}
+};
+var isDeviceLockedToolDefinition = {
+  name: "is_device_locked",
+  description: "checks if device is locked",
+  inputSchema: {}
+};
+var shakeDeviceToolDefinition = {
+  name: "shake_device",
+  description: "shakes the device (iOS only)",
+  inputSchema: {}
+};
+var hideKeyboardToolDefinition = {
+  name: "hide_keyboard",
+  description: "hides the on-screen keyboard",
+  inputSchema: {}
+};
+var isKeyboardShownToolDefinition = {
+  name: "is_keyboard_shown",
+  description: "checks if keyboard is visible",
+  inputSchema: {}
+};
+var openNotificationsToolDefinition = {
+  name: "open_notifications",
+  description: "opens the notifications panel (Android only)",
+  inputSchema: {}
+};
+var getGeolocationToolDefinition = {
+  name: "get_geolocation",
+  description: "gets current device geolocation",
+  inputSchema: {}
+};
+var rotateDeviceToolDefinition = {
+  name: "rotate_device",
+  description: "rotates device to portrait or landscape orientation",
+  inputSchema: {
+    orientation: z18.enum(["PORTRAIT", "LANDSCAPE"]).describe("Device orientation")
+  }
+};
+var sendKeysToolDefinition = {
+  name: "send_keys",
+  description: "sends keys to the app (Android only)",
+  inputSchema: {
+    keys: z18.array(z18.string()).describe('Array of keys to send (e.g., ["h", "e", "l", "l", "o"])')
+  }
+};
+var pressKeyCodeToolDefinition = {
+  name: "press_key_code",
+  description: "presses an Android key code (Android only)",
+  inputSchema: {
+    keyCode: z18.number().describe("Android key code (e.g., 4 for BACK, 3 for HOME)")
+  }
+};
+var setGeolocationToolDefinition = {
+  name: "set_geolocation",
+  description: "sets device geolocation (latitude, longitude, altitude)",
+  inputSchema: {
+    latitude: z18.number().min(-90).max(90).describe("Latitude coordinate"),
+    longitude: z18.number().min(-180).max(180).describe("Longitude coordinate"),
+    altitude: z18.number().optional().describe("Altitude in meters (optional)")
+  }
+};
 var getDeviceInfoTool = async () => {
   try {
     const browser = getBrowser();
@@ -1949,9 +2203,6 @@ ${Object.entries(info).map(([key, value]) => `  ${key}: ${value}`).join("\n")}`
     };
   }
 };
-var rotateDeviceToolArguments = {
-  orientation: z18.enum(["PORTRAIT", "LANDSCAPE"]).describe("Device orientation")
-};
 var rotateDeviceTool = async (args) => {
   try {
     const browser = getBrowser();
@@ -2031,9 +2282,6 @@ var shakeDeviceTool = async () => {
     };
   }
 };
-var sendKeysToolArguments = {
-  keys: z18.array(z18.string()).describe('Array of keys to send (e.g., ["h", "e", "l", "l", "o"])')
-};
 var sendKeysTool = async (args) => {
   try {
     const browser = getBrowser();
@@ -2048,9 +2296,6 @@ var sendKeysTool = async (args) => {
     };
   }
 };
-var pressKeyCodeToolArguments = {
-  keyCode: z18.number().describe("Android key code (e.g., 4 for BACK, 3 for HOME)")
-};
 var pressKeyCodeTool = async (args) => {
   try {
     const browser = getBrowser();
@@ -2125,11 +2370,6 @@ var getGeolocationTool = async () => {
     };
   }
 };
-var setGeolocationToolArguments = {
-  latitude: z18.number().min(-90).max(90).describe("Latitude coordinate"),
-  longitude: z18.number().min(-180).max(180).describe("Longitude coordinate"),
-  altitude: z18.number().optional().describe("Altitude in meters (optional)")
-};
 var setGeolocationTool = async (args) => {
   try {
     const browser = getBrowser();
@@ -2153,62 +2393,129 @@ var setGeolocationTool = async (args) => {
   }
 };
 
+// package.json
+var package_default = {
+  name: "@wdio/mcp",
+  author: "Vince Graics",
+  repository: {
+    type: "git",
+    url: "git://github.com/webdriverio/mcp.git"
+  },
+  version: "1.5.1",
+  description: "MCP server with WebdriverIO for browser and mobile app automation (iOS/Android via Appium)",
+  main: "./lib/server.js",
+  module: "./lib/server.js",
+  types: "./lib/server.d.ts",
+  bin: {
+    "wdio-mcp": "lib/server.js"
+  },
+  license: "MIT",
+  publishConfig: {
+    access: "public"
+  },
+  type: "module",
+  files: [
+    "lib",
+    "README.md"
+  ],
+  scripts: {
+    prebundle: "rimraf lib --glob ./*.tgz",
+    bundle: "tsup && shx chmod +x lib/server.js",
+    postbundle: "npm pack",
+    lint: "eslint src/ --fix && tsc --noEmit",
+    start: "node lib/server.js",
+    dev: "tsx --watch src/server.ts",
+    prepare: "husky"
+  },
+  dependencies: {
+    "@modelcontextprotocol/sdk": "1.25",
+    "@toon-format/toon": "^2.1.0",
+    "@wdio/protocols": "^9.16.2",
+    "@xmldom/xmldom": "^0.8.11",
+    "puppeteer-core": "^24.35.0",
+    webdriverio: "9.23",
+    zod: "^4.3.5"
+  },
+  devDependencies: {
+    "@release-it/conventional-changelog": "^10.0.4",
+    "@types/node": "^20.11.0",
+    "@wdio/eslint": "^0.1.3",
+    "@wdio/types": "^9.20.0",
+    eslint: "^9.39.2",
+    husky: "^9.1.7",
+    "release-it": "^19.2.3",
+    rimraf: "^6.1.2",
+    shx: "^0.4.0",
+    tsup: "^8.5.1",
+    tsx: "^4.21.0",
+    typescript: "5.9"
+  },
+  packageManager: "pnpm@10.12.4"
+};
+
 // src/server.ts
 console.log = (...args) => console.error("[LOG]", ...args);
 console.info = (...args) => console.error("[INFO]", ...args);
 console.warn = (...args) => console.error("[WARN]", ...args);
 console.debug = (...args) => console.error("[DEBUG]", ...args);
 var server = new McpServer({
-  name: "MCP WebdriverIO",
-  version: "1.4.0"
+  title: "WebdriverIO MCP Server",
+  name: package_default.name,
+  version: package_default.version,
+  description: package_default.description,
+  websiteUrl: "https://github.com/webdriverio/mcp"
 }, {
+  instructions: "MCP server for browser and mobile app automation using WebDriverIO. Supports Chrome browser control (headed/headless) and iOS/Android native app testing via Appium.",
   capabilities: {
-    resources: {},
     tools: {}
   }
 });
-server.tool("start_browser", "starts a browser session and sets it to the current state", startBrowserToolArguments, startBrowserTool);
-server.tool("start_app_session", "starts a mobile app session (iOS/Android) via Appium", startAppToolArguments, startAppTool);
-server.tool("close_session", "closes or detaches from the current browser or app session", closeSessionToolArguments, closeSessionTool);
-server.tool("navigate", "navigates to a URL", navigateToolArguments, navigateTool);
-server.tool("get_visible_elements", 'get a list of visible (in viewport & displayed) interactable elements on the page (buttons, links, inputs). Use elementType="visual" for images/SVGs. Must prefer this to take_screenshot for interactions', getVisibleElementsToolArguments, getVisibleElementsTool);
-server.tool("get_accessibility", "gets accessibility tree snapshot with semantic information about page elements (roles, names, states). Browser-only - use when get_visible_elements does not return expected elements.", getAccessibilityToolArguments, getAccessibilityTreeTool);
-server.tool("scroll_down", "scrolls the page down by specified pixels", scrollDownToolArguments, scrollDownTool);
-server.tool("scroll_up", "scrolls the page up by specified pixels", scrollUpToolArguments, scrollUpTool);
-server.tool("find_element", "finds an element", findElementToolArguments, findElementTool);
-server.tool("click_element", "clicks an element", clickToolArguments, clickTool);
-server.tool("click_via_text", "clicks an element", clickToolArguments, clickToolViaText);
-server.tool("set_value", "set value to an element, aka typing", setValueToolArguments, setValueTool);
-server.tool("get_element_text", "gets the text content of an element", getElementTextToolArguments, getElementTextTool);
-server.tool("is_displayed", "checks if an element is displayed", isDisplayedToolArguments, isDisplayedTool);
-server.tool("take_screenshot", "captures a screenshot of the current page", takeScreenshotToolArguments, takeScreenshotTool);
-server.tool("get_cookies", "gets all cookies or a specific cookie by name", getCookiesToolArguments, getCookiesTool);
-server.tool("set_cookie", "sets a cookie with specified name, value, and optional attributes", setCookieToolArguments, setCookieTool);
-server.tool("delete_cookies", "deletes all cookies or a specific cookie by name", deleteCookiesToolArguments, deleteCookiesTool);
-server.tool("tap_element", "taps an element by selector or coordinates (mobile)", tapElementToolArguments, tapElementTool);
-server.tool("swipe", "performs a swipe gesture in specified direction (mobile)", swipeToolArguments, swipeTool);
-server.tool("long_press", "performs a long press on element or coordinates (mobile)", longPressToolArguments, longPressTool);
-server.tool("drag_and_drop", "drags from one location to another (mobile)", dragAndDropToolArguments, dragAndDropTool);
-server.tool("get_app_state", "gets the state of an app (not installed, not running, background, foreground)", getAppStateToolArguments, getAppStateTool);
-server.tool("activate_app", "activates/brings an app to foreground", activateAppToolArguments, activateAppTool);
-server.tool("terminate_app", "terminates a running app", terminateAppToolArguments, terminateAppTool);
-server.tool("get_contexts", "lists available contexts (NATIVE_APP, WEBVIEW)", {}, getContextsTool);
-server.tool("get_current_context", "shows the currently active context", {}, getCurrentContextTool);
-server.tool("switch_context", "switches between native and webview contexts", switchContextToolArguments, switchContextTool);
-server.tool("get_device_info", "gets device information (platform, version, screen size)", {}, getDeviceInfoTool);
-server.tool("rotate_device", "rotates device to portrait or landscape orientation", rotateDeviceToolArguments, rotateDeviceTool);
-server.tool("get_orientation", "gets current device orientation", {}, getOrientationTool);
-server.tool("lock_device", "locks the device screen", {}, lockDeviceTool);
-server.tool("unlock_device", "unlocks the device screen", {}, unlockDeviceTool);
-server.tool("is_device_locked", "checks if device is locked", {}, isDeviceLockedTool);
-server.tool("shake_device", "shakes the device (iOS only)", {}, shakeDeviceTool);
-server.tool("send_keys", "sends keys to the app (Android only)", sendKeysToolArguments, sendKeysTool);
-server.tool("press_key_code", "presses an Android key code (Android only)", pressKeyCodeToolArguments, pressKeyCodeTool);
-server.tool("hide_keyboard", "hides the on-screen keyboard", {}, hideKeyboardTool);
-server.tool("is_keyboard_shown", "checks if keyboard is visible", {}, isKeyboardShownTool);
-server.tool("open_notifications", "opens the notifications panel (Android only)", {}, openNotificationsTool);
-server.tool("get_geolocation", "gets current device geolocation", {}, getGeolocationTool);
-server.tool("set_geolocation", "sets device geolocation (latitude, longitude, altitude)", setGeolocationToolArguments, setGeolocationTool);
+var registerTool = (definition, callback) => server.registerTool(definition.name, {
+  description: definition.description,
+  inputSchema: definition.inputSchema
+}, callback);
+registerTool(startBrowserToolDefinition, startBrowserTool);
+registerTool(startAppToolDefinition, startAppTool);
+registerTool(closeSessionToolDefinition, closeSessionTool);
+registerTool(navigateToolDefinition, navigateTool);
+registerTool(getVisibleElementsToolDefinition, getVisibleElementsTool);
+registerTool(getAccessibilityToolDefinition, getAccessibilityTreeTool);
+registerTool(scrollDownToolDefinition, scrollDownTool);
+registerTool(scrollUpToolDefinition, scrollUpTool);
+registerTool(findElementToolDefinition, findElementTool);
+registerTool(clickToolDefinition, clickTool);
+registerTool(clickViaTextToolDefinition, clickToolViaText);
+registerTool(setValueToolDefinition, setValueTool);
+registerTool(getElementTextToolDefinition, getElementTextTool);
+registerTool(isDisplayedToolDefinition, isDisplayedTool);
+registerTool(takeScreenshotToolDefinition, takeScreenshotTool);
+registerTool(getCookiesToolDefinition, getCookiesTool);
+registerTool(setCookieToolDefinition, setCookieTool);
+registerTool(deleteCookiesToolDefinition, deleteCookiesTool);
+registerTool(tapElementToolDefinition, tapElementTool);
+registerTool(swipeToolDefinition, swipeTool);
+registerTool(longPressToolDefinition, longPressTool);
+registerTool(dragAndDropToolDefinition, dragAndDropTool);
+registerTool(getAppStateToolDefinition, getAppStateTool);
+registerTool(activateAppToolDefinition, activateAppTool);
+registerTool(terminateAppToolDefinition, terminateAppTool);
+registerTool(getContextsToolDefinition, getContextsTool);
+registerTool(getCurrentContextToolDefinition, getCurrentContextTool);
+registerTool(switchContextToolDefinition, switchContextTool);
+registerTool(getDeviceInfoToolDefinition, getDeviceInfoTool);
+registerTool(rotateDeviceToolDefinition, rotateDeviceTool);
+registerTool(getOrientationToolDefinition, getOrientationTool);
+registerTool(lockDeviceToolDefinition, lockDeviceTool);
+registerTool(unlockDeviceToolDefinition, unlockDeviceTool);
+registerTool(isDeviceLockedToolDefinition, isDeviceLockedTool);
+registerTool(shakeDeviceToolDefinition, shakeDeviceTool);
+registerTool(sendKeysToolDefinition, sendKeysTool);
+registerTool(pressKeyCodeToolDefinition, pressKeyCodeTool);
+registerTool(hideKeyboardToolDefinition, hideKeyboardTool);
+registerTool(isKeyboardShownToolDefinition, isKeyboardShownTool);
+registerTool(openNotificationsToolDefinition, openNotificationsTool);
+registerTool(getGeolocationToolDefinition, getGeolocationTool);
+registerTool(setGeolocationToolDefinition, setGeolocationTool);
 async function main() {
   const transport = new StdioServerTransport();
   await server.connect(transport);