@testdriverai/mcp 7.11.177-test → 7.11.178-test

package/mcp-server/dist/server.mjs

@@ -511,6 +511,7 @@ function buildServer() {
     // Session Start
     registerAppTool(server, "session_start", {
         title: "Session Start",
+        annotations: { title: "Session Start", readOnlyHint: false, destructiveHint: true },
         description: `Start a new TestDriver session and provision a sandbox with browser or app.
 
 ⚠️ IMPORTANT - Test File Parameter:
@@ -644,6 +645,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     });
     // Session Status
     server.registerTool("session_status", {
+        annotations: { title: "Session Status", readOnlyHint: true },
         description: "Check the current session status and time remaining",
         inputSchema: z.object({}),
     }, async () => {
@@ -661,6 +663,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     });
     // Session Extend
     server.registerTool("session_extend", {
+        annotations: { title: "Session Extend", readOnlyHint: false, destructiveHint: false },
         description: "Extend the session keepAlive time",
         inputSchema: z.object({
             additionalMs: z.number().default(60000).describe("Additional time in ms"),
@@ -686,6 +689,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     // Find Element
     registerAppTool(server, "find", {
         title: "Find Element",
+        annotations: { title: "Find Element", readOnlyHint: true },
         description: "Find an element on screen by natural language description",
         inputSchema: z.object({
             description: z.string().describe("Natural language description of the element"),
@@ -726,6 +730,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     // Find All Elements
     registerAppTool(server, "findall", {
         title: "Find All Elements",
+        annotations: { title: "Find All Elements", readOnlyHint: true },
         description: "Find all elements on screen matching a natural language description. Returns an array of element references.",
         inputSchema: z.object({
             description: z.string().describe("Natural language description of the elements to find"),
@@ -766,6 +771,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     // Click
     registerAppTool(server, "click", {
         title: "Click Element",
+        annotations: { title: "Click Element", readOnlyHint: false, destructiveHint: true },
         description: "Click on a previously found element. Use 'find' first to locate the element.",
         inputSchema: z.object({
             elementRef: z.string().describe("Reference to previously found element (required). Get this from a 'find' call."),
@@ -795,6 +801,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     // Hover
     registerAppTool(server, "hover", {
         title: "Hover Element",
+        annotations: { title: "Hover Element", readOnlyHint: true },
         description: "Hover over a previously found element. Use 'find' first to locate the element.",
         inputSchema: z.object({
             elementRef: z.string().describe("Reference to previously found element (required). Get this from a 'find' call."),
@@ -822,6 +829,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     });
     // Wait
     server.registerTool("wait", {
+        annotations: { title: "Wait", readOnlyHint: true },
         description: "Wait for a specified amount of time",
         inputSchema: z.object({
             timeout: z.number().default(3000).describe("Time to wait in milliseconds (default: 3000)"),
@@ -848,6 +856,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     });
     // Focus Application
     server.registerTool("focus_application", {
+        annotations: { title: "Focus Application", readOnlyHint: false, destructiveHint: true },
         description: "Bring an application window to the foreground",
         inputSchema: z.object({
             name: z.string().describe("Name of the application to focus (e.g., 'Google Chrome', 'Visual Studio Code')"),
@@ -875,6 +884,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     // Find and Click
     registerAppTool(server, "find_and_click", {
         title: "Find and Click",
+        annotations: { title: "Find and Click", readOnlyHint: false, destructiveHint: true },
         description: "Find an element and click it in one action",
         inputSchema: z.object({
             description: z.string().describe("Natural language description of element"),
@@ -914,6 +924,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     });
     // Type
     server.registerTool("type", {
+        annotations: { title: "Type Text", readOnlyHint: false, destructiveHint: true },
         description: "Type text into the currently focused field",
         inputSchema: z.object({
             text: z.string().describe("Text to type"),
@@ -942,6 +953,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     });
     // Press Keys
     server.registerTool("press_keys", {
+        annotations: { title: "Press Keys", readOnlyHint: false, destructiveHint: true },
         description: "Press keyboard keys or shortcuts",
         inputSchema: z.object({
             keys: z.array(z.string()).describe("Array of keys to press (e.g., ['ctrl', 'a'])"),
@@ -968,6 +980,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     });
     // Scroll
     server.registerTool("scroll", {
+        annotations: { title: "Scroll", readOnlyHint: false, destructiveHint: true },
         description: "Scroll the page or element",
         inputSchema: z.object({
             direction: z.enum(["up", "down", "left", "right"]).default("down"),
@@ -995,6 +1008,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
     });
     // Assert - generates code for test files
     server.registerTool("assert", {
+        annotations: { title: "Assert", readOnlyHint: true },
         description: `Make an AI-powered assertion about the current screen state. GENERATES CODE for the test file.
 
 Use this when you want a verification step recorded in the generated test. This will add code like:
@@ -1028,6 +1042,7 @@ Unlike 'check' which is for your understanding during development, 'assert' crea
     // Check - AI uses this to understand the screen state (DOES NOT generate code)
     registerAppTool(server, "check", {
         title: "Check Screen State",
+        annotations: { title: "Check Screen State", readOnlyHint: true },
         description: `👁️ THIS IS HOW YOU SEE THE SCREEN. Use this tool whenever you need to understand what's currently displayed.
 
 This tool captures a screenshot and returns AI analysis to YOU. Use it to:
@@ -1116,6 +1131,7 @@ You can optionally provide a reference image URI to compare against a previous s
     });
     // Exec
     server.registerTool("exec", {
+        annotations: { title: "Execute Command", readOnlyHint: false, destructiveHint: true },
         description: "Execute shell or PowerShell commands in the sandbox",
         inputSchema: z.object({
             language: z.enum(["sh", "pwsh"]).default("sh"),
@@ -1158,6 +1174,7 @@ You can optionally provide a reference image URI to compare against a previous s
     }
     // List Local Screenshots - lists screenshots saved to .testdriver directory
     server.registerTool("list_local_screenshots", {
+        annotations: { title: "List Local Screenshots", readOnlyHint: true },
         description: `List and filter screenshots saved in the .testdriver directory.
 
 Screenshots from auto-screenshot feature use the format: <seq>-<action>-<phase>-L<line>-<description>.png
@@ -1378,6 +1395,7 @@ Returns a list of screenshot paths that can be viewed with the 'view_local_scree
     // Also displays to the user via MCP App
     registerAppTool(server, "view_local_screenshot", {
         title: "View Local Screenshot",
+        annotations: { title: "View Local Screenshot", readOnlyHint: true },
         description: `View a screenshot from the .testdriver directory.
 
 Use 'list_local_screenshots' first to see available screenshots, then use this tool to view one.
@@ -1457,6 +1475,7 @@ Useful for:
     // Use 'check' tool for AI to understand screen state.
     registerAppTool(server, "screenshot", {
         title: "Screenshot",
+        annotations: { title: "Screenshot", readOnlyHint: true },
         description: `Display a screenshot to the user. This tool does NOT return the image to you (the AI).
 
 ⚠️ IMPORTANT: Do NOT use this tool to understand the screen state. The screenshot is ONLY displayed to the human user - you will NOT receive the image or any analysis.
@@ -1497,6 +1516,7 @@ Only use 'screenshot' when you explicitly want to show something to the human us
     });
     // Init - Initialize a new TestDriver project
     server.registerTool("init", {
+        annotations: { title: "Initialize Project", readOnlyHint: false, destructiveHint: true },
         description: `Initialize a new TestDriver project with Vitest SDK examples.
 
 This creates:
@@ -1576,6 +1596,7 @@ Learn more at https://docs.testdriver.ai/v7/getting-started/
     // the caller's team by their OAuth token (hosted mode) or env API key.
     // =============================================================================
     server.registerTool("list_test_runs", {
+        annotations: { title: "List Test Runs", readOnlyHint: true },
         description: "List recent TestDriver test runs for the current team.",
         inputSchema: z.object({
             page: z.number().optional().describe("Page number (1-indexed)"),
@@ -1598,6 +1619,7 @@ Learn more at https://docs.testdriver.ai/v7/getting-started/
         }
     });
     server.registerTool("get_test_run_detail", {
+        annotations: { title: "Get Test Run Detail", readOnlyHint: true },
         description: "Get details for a single TestDriver test run, including its test cases.",
         inputSchema: z.object({
             id: z.string().describe("The test run id"),

package/mcp-server/mcp-app.html

@@ -11,7 +11,7 @@
     <div class="screenshot-wrapper">
       <div class="loading-overlay" id="loading-overlay">
         <div class="loading-spinner"></div>
-        <span class="loading-text">Waiting for screenshot...</span>
+        <span class="loading-text">Connecting to session...</span>
       </div>
       <div class="screenshot-container" id="screenshot-container" style="display: none;">
         <img id="screenshot" alt="Screenshot" />

package/mcp-server/src/mcp-app.css

@@ -62,7 +62,7 @@ body {
   width: 32px;
   height: 32px;
   border: 3px solid var(--color-border-primary, #e5e5e5);
-  border-top-color: #3b82f6;
+  border-top-color: #b3d334;
   border-radius: 50%;
   animation: spin 0.8s linear infinite;
 }
@@ -99,8 +99,8 @@ body {
   position: absolute;
   top: 0;
   left: 0;
-  width: 100%;
-  height: 100%;
+  width: 0;
+  height: 0;
   pointer-events: none;
 }
 
@@ -231,7 +231,6 @@ body {
   align-items: center;
   padding: 8px 12px;
   background: var(--color-background-secondary, #f5f5f5);
-  border-top: 1px solid var(--color-border-primary, #e5e5e5);
   font-size: var(--font-text-xs-size, 12px);
 }
 
@@ -265,7 +264,6 @@ body {
 .target-info {
   padding: 8px 12px;
   background: var(--color-background-secondary, #f5f5f5);
-  border-bottom: 1px solid var(--color-border-primary, #e5e5e5);
   font-size: var(--font-text-sm-size, 14px);
   color: var(--color-text-primary, #1a1a1a);
 }

package/mcp-server/src/mcp-app.ts

@@ -219,6 +219,12 @@ function addOverlays(data: ToolResultData) {
       setTimeout(() => addOverlays(data), 50);
       return;
     }
+
+    // Align overlay layer to the image box so coordinates stay accurate when centered
+    overlaysEl.style.left = `${screenshotEl.offsetLeft}px`;
+    overlaysEl.style.top = `${screenshotEl.offsetTop}px`;
+    overlaysEl.style.width = `${displayedWidth}px`;
+    overlaysEl.style.height = `${displayedHeight}px`;
     
     // Track the focal point for zoom
     let focalX: number | undefined;
@@ -384,15 +390,25 @@ function renderResult(data: ToolResultData) {
   // Update target info for find/find_and_click actions
   if (data.element && (data.action === "find" || data.action === "find_and_click")) {
     const el = data.element;
-    let targetHtml = `<strong>Target:</strong> "${el.description || "Element"}"`;
+    targetInfoEl.replaceChildren();
+
+    const labelEl = document.createElement("strong");
+    labelEl.textContent = "Target:";
+    targetInfoEl.append(labelEl, ` ${el.description || "Element"}`);
+
     if (el.centerX !== undefined && el.centerY !== undefined) {
-      targetHtml += ` <span class="target-coords">(${Math.round(el.centerX)}, ${Math.round(el.centerY)})</span>`;
+      const coordsEl = document.createElement("span");
+      coordsEl.className = "target-coords";
+      coordsEl.textContent = `(${Math.round(el.centerX)}, ${Math.round(el.centerY)})`;
+      targetInfoEl.append(" ", coordsEl);
     }
     if (el.confidence !== undefined) {
       const confidencePercent = Math.round(el.confidence * 100);
-      targetHtml += ` <span class="target-confidence ${confidencePercent >= 70 ? 'high' : confidencePercent >= 40 ? 'medium' : 'low'}">${confidencePercent}%</span>`;
+      const confidenceEl = document.createElement("span");
+      confidenceEl.className = `target-confidence ${confidencePercent >= 70 ? 'high' : confidencePercent >= 40 ? 'medium' : 'low'}`;
+      confidenceEl.textContent = `${confidencePercent}%`;
+      targetInfoEl.append(" ", confidenceEl);
     }
-    targetInfoEl.innerHTML = targetHtml;
     targetInfoEl.classList.remove("hidden");
   } else {
     targetInfoEl.classList.add("hidden");
@@ -413,7 +429,7 @@ function renderResult(data: ToolResultData) {
       screenshotNaturalWidth = screenshotEl.naturalWidth;
       screenshotNaturalHeight = screenshotEl.naturalHeight;
       // Show the container and add overlays now that we know dimensions
-      containerEl.style.display = "block";
+      containerEl.style.display = "flex";
       addOverlays(data);
       // Hide loading state
       hideLoading();
@@ -531,11 +547,16 @@ app.onerror = (error) => {
 app.onhostcontextchanged = handleHostContextChanged;
 
 // 3. Connect to host
+showLoading("Connecting to session...");
 app.connect().then(() => {
+  showLoading("Waiting for first action...");
   const ctx = app.getHostContext();
   if (ctx) {
     handleHostContextChanged(ctx);
   }
+}).catch((error) => {
+  console.error("Failed to connect to host:", error);
+  showLoading("Connection failed");
 });
 
 // Insert target info element after screenshot wrapper

package/mcp-server/src/server.ts

@@ -672,6 +672,7 @@ registerAppTool(
   "session_start",
   {
     title: "Session Start",
+    annotations: { title: "Session Start", readOnlyHint: false, destructiveHint: true },
     description: `Start a new TestDriver session and provision a sandbox with browser or app.
 
 ⚠️ IMPORTANT - Test File Parameter:
@@ -825,6 +826,7 @@ IMPORTANT - If creating a new test project, use these EXACT dependencies in pack
 server.registerTool(
   "session_status",
   {
+    annotations: { title: "Session Status", readOnlyHint: true },
     description: "Check the current session status and time remaining",
     inputSchema: z.object({}),
   },
@@ -849,6 +851,7 @@ server.registerTool(
 server.registerTool(
   "session_extend",
   {
+    annotations: { title: "Session Extend", readOnlyHint: false, destructiveHint: false },
     description: "Extend the session keepAlive time",
     inputSchema: z.object({
       additionalMs: z.number().default(60000).describe("Additional time in ms"),
@@ -884,6 +887,7 @@ registerAppTool(
   "find",
   {
     title: "Find Element",
+    annotations: { title: "Find Element", readOnlyHint: true },
     description: "Find an element on screen by natural language description",
     inputSchema: z.object({
       description: z.string().describe("Natural language description of the element"),
@@ -930,6 +934,7 @@ registerAppTool(
   "findall",
   {
     title: "Find All Elements",
+    annotations: { title: "Find All Elements", readOnlyHint: true },
     description: "Find all elements on screen matching a natural language description. Returns an array of element references.",
     inputSchema: z.object({
       description: z.string().describe("Natural language description of the elements to find"),
@@ -976,6 +981,7 @@ registerAppTool(
   "click",
   {
     title: "Click Element",
+    annotations: { title: "Click Element", readOnlyHint: false, destructiveHint: true },
     description: "Click on a previously found element. Use 'find' first to locate the element.",
     inputSchema: z.object({
       elementRef: z.string().describe("Reference to previously found element (required). Get this from a 'find' call."),
@@ -1013,6 +1019,7 @@ registerAppTool(
   "hover",
   {
     title: "Hover Element",
+    annotations: { title: "Hover Element", readOnlyHint: true },
     description: "Hover over a previously found element. Use 'find' first to locate the element.",
     inputSchema: z.object({
       elementRef: z.string().describe("Reference to previously found element (required). Get this from a 'find' call."),
@@ -1047,6 +1054,7 @@ registerAppTool(
 server.registerTool(
   "wait",
   {
+    annotations: { title: "Wait", readOnlyHint: true },
     description: "Wait for a specified amount of time",
     inputSchema: z.object({
       timeout: z.number().default(3000).describe("Time to wait in milliseconds (default: 3000)"),
@@ -1080,6 +1088,7 @@ server.registerTool(
 server.registerTool(
   "focus_application",
   {
+    annotations: { title: "Focus Application", readOnlyHint: false, destructiveHint: true },
     description: "Bring an application window to the foreground",
     inputSchema: z.object({
       name: z.string().describe("Name of the application to focus (e.g., 'Google Chrome', 'Visual Studio Code')"),
@@ -1115,6 +1124,7 @@ registerAppTool(
   "find_and_click",
   {
     title: "Find and Click",
+    annotations: { title: "Find and Click", readOnlyHint: false, destructiveHint: true },
     description: "Find an element and click it in one action",
     inputSchema: z.object({
       description: z.string().describe("Natural language description of element"),
@@ -1159,6 +1169,7 @@ registerAppTool(
 server.registerTool(
   "type",
   {
+    annotations: { title: "Type Text", readOnlyHint: false, destructiveHint: true },
     description: "Type text into the currently focused field",
     inputSchema: z.object({
       text: z.string().describe("Text to type"),
@@ -1194,6 +1205,7 @@ server.registerTool(
 server.registerTool(
   "press_keys",
   {
+    annotations: { title: "Press Keys", readOnlyHint: false, destructiveHint: true },
     description: "Press keyboard keys or shortcuts",
     inputSchema: z.object({
       keys: z.array(z.string()).describe("Array of keys to press (e.g., ['ctrl', 'a'])"),
@@ -1227,6 +1239,7 @@ server.registerTool(
 server.registerTool(
   "scroll",
   {
+    annotations: { title: "Scroll", readOnlyHint: false, destructiveHint: true },
     description: "Scroll the page or element",
     inputSchema: z.object({
       direction: z.enum(["up", "down", "left", "right"]).default("down"),
@@ -1261,6 +1274,7 @@ server.registerTool(
 server.registerTool(
   "assert",
   {
+    annotations: { title: "Assert", readOnlyHint: true },
     description: `Make an AI-powered assertion about the current screen state. GENERATES CODE for the test file.
 
 Use this when you want a verification step recorded in the generated test. This will add code like:
@@ -1302,6 +1316,7 @@ registerAppTool(
   "check",
   {
     title: "Check Screen State",
+    annotations: { title: "Check Screen State", readOnlyHint: true },
     description: `👁️ THIS IS HOW YOU SEE THE SCREEN. Use this tool whenever you need to understand what's currently displayed.
 
 This tool captures a screenshot and returns AI analysis to YOU. Use it to:
@@ -1398,6 +1413,7 @@ You can optionally provide a reference image URI to compare against a previous s
 server.registerTool(
   "exec",
   {
+    annotations: { title: "Execute Command", readOnlyHint: false, destructiveHint: true },
     description: "Execute shell or PowerShell commands in the sandbox",
     inputSchema: z.object({
       language: z.enum(["sh", "pwsh"]).default("sh"),
@@ -1459,6 +1475,7 @@ function parseScreenshotFilename(filename: string): ParsedScreenshotInfo {
 server.registerTool(
   "list_local_screenshots",
   {
+    annotations: { title: "List Local Screenshots", readOnlyHint: true },
     description: `List and filter screenshots saved in the .testdriver directory.
 
 Screenshots from auto-screenshot feature use the format: <seq>-<action>-<phase>-L<line>-<description>.png
@@ -1691,6 +1708,7 @@ registerAppTool(
   "view_local_screenshot",
   {
     title: "View Local Screenshot",
+    annotations: { title: "View Local Screenshot", readOnlyHint: true },
     description: `View a screenshot from the .testdriver directory.
 
 Use 'list_local_screenshots' first to see available screenshots, then use this tool to view one.
@@ -1784,6 +1802,7 @@ registerAppTool(
   "screenshot",
   {
     title: "Screenshot",
+    annotations: { title: "Screenshot", readOnlyHint: true },
     description: `Display a screenshot to the user. This tool does NOT return the image to you (the AI).
 
 ⚠️ IMPORTANT: Do NOT use this tool to understand the screen state. The screenshot is ONLY displayed to the human user - you will NOT receive the image or any analysis.
@@ -1832,6 +1851,7 @@ Only use 'screenshot' when you explicitly want to show something to the human us
 server.registerTool(
   "init",
   {
+    annotations: { title: "Initialize Project", readOnlyHint: false, destructiveHint: true },
     description: `Initialize a new TestDriver project with Vitest SDK examples.
 
 This creates:
@@ -1927,6 +1947,7 @@ Learn more at https://docs.testdriver.ai/v7/getting-started/
 server.registerTool(
   "list_test_runs",
   {
+    annotations: { title: "List Test Runs", readOnlyHint: true },
     description: "List recent TestDriver test runs for the current team.",
     inputSchema: z.object({
       page: z.number().optional().describe("Page number (1-indexed)"),
@@ -1953,6 +1974,7 @@ server.registerTool(
 server.registerTool(
   "get_test_run_detail",
   {
+    annotations: { title: "Get Test Run Detail", readOnlyHint: true },
     description: "Get details for a single TestDriver test run, including its test cases.",
     inputSchema: z.object({
       id: z.string().describe("The test run id"),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@testdriverai/mcp",
-  "version": "7.11.177-test",
+  "version": "7.11.178-test",
   "description": "Next generation autonomous AI agent for end-to-end testing of web & desktop",
   "main": "sdk.js",
   "types": "sdk.d.ts",