safari-mcp 2.4.0 → 2.5.1

package/extension/background.js

@@ -405,7 +405,7 @@ async function handleCommand(type, payload) {
 
     // --- Click & Input ---
     case "click": {
-      return await execInTab((selector, text, x, y, ref) => {
+      const result = await execInTab((selector, text, x, y, ref) => {
         // Use shared deep query (defined by ensureHelpers / _deepQueryScript)
         const dq = window.__mcpDeepQuery || document.querySelector.bind(document);
 
@@ -633,6 +633,36 @@ async function handleCommand(type, payload) {
 
         return "Clicked: " + el.tagName + (el.textContent ? ' "' + el.textContent.trim().substring(0, 50) + '"' : "");
       }, [payload.selector, payload.text, payload.x, payload.y, payload.ref], tabId);
+
+      // Fallback: if element not found in main frame, try all frames (cross-origin iframes)
+      if (result && (result.startsWith("Element not found") || result === "No click target")) {
+        const iframeResult = await execInAllFrames((selector, text) => {
+          let el = null;
+          if (selector) {
+            el = document.querySelector(selector);
+          } else if (text) {
+            // Search interactive elements by text
+            const candidates = document.querySelectorAll("button, a, [role='button'], input[type='submit']");
+            for (let i = 0; i < candidates.length; i++) {
+              const t = (candidates[i].innerText || candidates[i].textContent || "").trim();
+              if (t === text) { el = candidates[i]; break; }
+            }
+            // Fuzzy: contains match
+            if (!el) {
+              for (let i = 0; i < candidates.length; i++) {
+                const t = (candidates[i].innerText || candidates[i].textContent || "").trim();
+                if (t.includes(text) || text.includes(t)) { el = candidates[i]; break; }
+              }
+            }
+          }
+          if (!el) return null;
+          el.scrollIntoView({ block: "center", behavior: "instant" });
+          el.click();
+          return "Clicked (iframe): " + el.tagName + (el.textContent ? ' "' + el.textContent.trim().substring(0, 50) + '"' : "");
+        }, [payload.selector, payload.text], tabId);
+        if (iframeResult) return iframeResult;
+      }
+      return result;
     }
 
     // --- Click + Read (combo — saves 1 full MCP round-trip) ---

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "safari-mcp",
-  "version": "2.4.0",
+  "version": "2.5.1",
   "mcpName": "io.github.achiya-automation/safari-mcp",
   "description": "Safari browser automation for AI agents — native macOS, zero Chrome overhead. 80 tools via AppleScript + JavaScript.",
   "type": "module",

package/safari-helper.swift

@@ -128,6 +128,57 @@ func performNativeClick(x: Double, y: Double, doubleClick: Bool = false, windowI
   return ["result": "clicked at (\(Int(x)),\(Int(y)))\(doubleClick ? " (double)" : "")\(targetInfo)"]
 }
 
+// ========== CGEvent Native Keyboard ==========
+// Sends keyboard events (keystrokes, shortcuts) targeted to a specific window
+// WITHOUT activating Safari or stealing focus. Same principle as native click.
+
+func performNativeKeyboard(keyCode: UInt16, flags: CGEventFlags = [], windowId: Int64 = 0) -> [String: Any] {
+  // Get Safari PID
+  var safariPID: pid_t = 0
+  if windowId > 0 {
+    let ws = NSWorkspace.shared
+    for app in ws.runningApplications {
+      if app.bundleIdentifier == "com.apple.Safari" {
+        safariPID = app.processIdentifier
+        break
+      }
+    }
+    if safariPID == 0 {
+      return ["error": "Safari process not found"]
+    }
+  }
+
+  let kWindowField = CGEventField(rawValue: 91)!
+  let kWindowHandlerField = CGEventField(rawValue: 92)!
+
+  guard let keyDown = CGEvent(keyboardEventSource: nil, virtualKey: keyCode, keyDown: true) else {
+    return ["error": "Failed to create key down event"]
+  }
+  guard let keyUp = CGEvent(keyboardEventSource: nil, virtualKey: keyCode, keyDown: false) else {
+    return ["error": "Failed to create key up event"]
+  }
+
+  keyDown.flags = flags
+  keyUp.flags = flags
+
+  if windowId > 0 {
+    keyDown.setIntegerValueField(kWindowField, value: windowId)
+    keyDown.setIntegerValueField(kWindowHandlerField, value: windowId)
+    keyUp.setIntegerValueField(kWindowField, value: windowId)
+    keyUp.setIntegerValueField(kWindowHandlerField, value: windowId)
+    keyDown.postToPid(safariPID)
+    usleep(30_000)
+    keyUp.postToPid(safariPID)
+  } else {
+    keyDown.post(tap: .cghidEventTap)
+    usleep(30_000)
+    keyUp.post(tap: .cghidEventTap)
+  }
+
+  let targetInfo = windowId > 0 ? " (window \(windowId), background)" : ""
+  return ["result": "key \(keyCode) pressed\(targetInfo)"]
+}
+
 // ========== Response Helper ==========
 
 func respond(_ obj: [String: Any]) {
@@ -166,6 +217,24 @@ if args.count >= 4 && args[1] == "--click" {
   exit(result["error"] != nil ? 1 : 0)
 }
 
+// ========== CLI Mode: --paste --window WID ==========
+// Convenience: Cmd+V via CGEvent (no activate needed)
+if args.count >= 2 && args[1] == "--paste" {
+  var windowId: Int64 = 0
+  var i = 2
+  while i < args.count {
+    if args[i] == "--window" && i + 1 < args.count {
+      windowId = Int64(args[i + 1]) ?? 0
+      i += 1
+    }
+    i += 1
+  }
+  // keyCode 9 = V key
+  let result = performNativeKeyboard(keyCode: 9, flags: .maskCommand, windowId: windowId)
+  respond(result)
+  exit(result["error"] != nil ? 1 : 0)
+}
+
 // ========== Daemon Mode: JSON lines on stdin ==========
 
 while let line = readLine(strippingNewline: true) {
@@ -187,9 +256,31 @@ while let line = readLine(strippingNewline: true) {
     continue
   }
 
+  // Handle CGEvent keyboard command
+  // {"keyboard": {"keyCode": 9, "flags": ["cmd"], "windowId": 4127}}
+  // keyCode 9 = V, flags: cmd/shift/alt/ctrl
+  if let kbData = json["keyboard"] as? [String: Any],
+     let keyCode = kbData["keyCode"] as? Int {
+    let windowId = Int64((kbData["windowId"] as? Int) ?? 0)
+    var flags: CGEventFlags = []
+    if let flagNames = kbData["flags"] as? [String] {
+      for f in flagNames {
+        switch f.lowercased() {
+          case "cmd": flags.insert(.maskCommand)
+          case "shift": flags.insert(.maskShift)
+          case "alt", "option": flags.insert(.maskAlternate)
+          case "ctrl", "control": flags.insert(.maskControl)
+          default: break
+        }
+      }
+    }
+    respond(performNativeKeyboard(keyCode: UInt16(keyCode), flags: flags, windowId: windowId))
+    continue
+  }
+
   // Handle AppleScript command
   guard let script = json["script"] as? String else {
-    respond(["error": "invalid input — expected 'script' or 'click'"])
+    respond(["error": "invalid input — expected 'script', 'click', or 'keyboard'"])
     continue
   }
 

package/safari.js

@@ -420,6 +420,44 @@ function _helperNativeClick(x, y, doubleClick = false, windowId = 0, timeout = 5
   });
 }
 
+// Sends a CGEvent keyboard command to the Swift helper daemon.
+// No focus stealing — sends key events directly to the target window via PID.
+function _helperNativeKeyboard(keyCode, flags = [], windowId = 0, timeout = 5000) {
+  return new Promise((resolve, reject) => {
+    if (!_helperProc) startHelper();
+    if (!_helperProc || !_helperProc.stdin || !_helperProc.stdin.writable) {
+      reject(new Error("safari-helper not available for native keyboard"));
+      return;
+    }
+    let resolved = false;
+    const timer = setTimeout(() => {
+      if (resolved) return;
+      resolved = true;
+      const idx = _helperQueue.indexOf(cb);
+      if (idx >= 0) _helperQueue.splice(idx, 1);
+      reject(new Error("native keyboard timeout"));
+    }, timeout);
+
+    function cb(line) {
+      if (resolved) return;
+      resolved = true;
+      clearTimeout(timer);
+      try {
+        const parsed = JSON.parse(line);
+        if (parsed.error) reject(new Error(parsed.error));
+        else resolve(parsed.result ?? "");
+      } catch {
+        resolve(line);
+      }
+    }
+
+    _helperQueue.push(cb);
+    const cmd = { keyboard: { keyCode, flags } };
+    if (windowId) cmd.keyboard.windowId = windowId;
+    _helperProc.stdin.write(JSON.stringify(cmd) + "\n");
+  });
+}
+
 // Get Safari window bounds, toolbar height, and window ID for coordinate calculation
 async function _getSafariWindowGeometry() {
   await refreshTargetWindow();
@@ -1070,6 +1108,19 @@ export async function pressKey({ key, modifiers = [] }) {
 
     // Cmd+V (paste) — read clipboard via AppleScript (no activate!), inject via JS
     if (k === "v") {
+      // Cross-origin iframe: JS can't paste into it, use CGEvent Cmd+V (no focus steal)
+      const activeTag = await runJS(`document.activeElement ? document.activeElement.tagName : ''`);
+      if (activeTag === 'IFRAME') {
+        try {
+          const geo = await _getSafariWindowGeometry();
+          await _helperNativeKeyboard(9, ["cmd"], geo.windowId);
+        } catch {
+          await _helperNativeKeyboard(9, ["cmd"], 0);
+        }
+        await new Promise(r => setTimeout(r, 300));
+        return `Pressed: ${modifiers.join("+")}+v (CGEvent Cmd+V into iframe, no focus steal)`;
+      }
+
       const clipText = await osascript(`the clipboard as text`).catch(() => "");
       if (clipText) {
         const escaped = clipText.replace(/\\/g, "\\\\").replace(/'/g, "\\'").replace(/\n/g, "\\n").replace(/\r/g, "\\r").replace(/\t/g, "\\t");
@@ -1153,6 +1204,61 @@ export async function pressKey({ key, modifiers = [] }) {
   return `Pressed: ${modifiers.length ? modifiers.join("+") + "+" : ""}${key}`;
 }
 
+// Native typing via clipboard paste — for cross-origin iframes (Intercom, Zendesk, etc.)
+// JS can't access content inside cross-origin iframes, so we use OS-level CGEvent:
+// 1. Save current clipboard
+// 2. Set clipboard to our text
+// 3. Cmd+V via CGEvent keyboard (targeted to window — NO focus steal, NO activate)
+// 4. Restore clipboard
+// Requires nativeClick to have focused the input inside the iframe first.
+async function _nativeTypeViaClipboard(text) {
+  // Save current clipboard
+  let savedClipboard = null;
+  try {
+    const { stdout } = await execFileAsync("pbpaste", []);
+    savedClipboard = stdout;
+  } catch {}
+
+  // Set clipboard to our text via pipe (safe from shell injection)
+  await new Promise((resolve, reject) => {
+    const proc = spawn("pbcopy", [], { stdio: ["pipe", "ignore", "ignore"] });
+    proc.stdin.write(text);
+    proc.stdin.end();
+    proc.on("close", resolve);
+    proc.on("error", reject);
+  });
+
+  // Paste via CGEvent Cmd+V targeted to Safari window — NO activate, NO focus steal
+  try {
+    const geo = await _getSafariWindowGeometry();
+    // keyCode 9 = V key, flags: ["cmd"]
+    await _helperNativeKeyboard(9, ["cmd"], geo.windowId);
+  } catch {
+    // Fallback: try without windowId (global)
+    await _helperNativeKeyboard(9, ["cmd"], 0);
+  }
+
+  // Wait for paste to settle
+  await new Promise(r => setTimeout(r, 300));
+
+  // Restore clipboard after delay
+  if (savedClipboard !== null) {
+    setTimeout(async () => {
+      try {
+        await new Promise((resolve, reject) => {
+          const proc = spawn("pbcopy", [], { stdio: ["pipe", "ignore", "ignore"] });
+          proc.stdin.write(savedClipboard);
+          proc.stdin.end();
+          proc.on("close", resolve);
+          proc.on("error", reject);
+        });
+      } catch {}
+    }, 3000);
+  }
+
+  return `Typed ${text.length} chars (native paste into iframe)`;
+}
+
 export async function typeText({ text, selector, ref }) {
   if (ref) selector = refSelector(ref);
   if (selector) {
@@ -1162,6 +1268,13 @@ export async function typeText({ text, selector, ref }) {
     await new Promise((r) => setTimeout(r, 30));
   }
 
+  // Cross-origin iframe detection: JS can't access content inside cross-origin iframes.
+  // When activeElement is an IFRAME, use native clipboard paste via System Events.
+  const activeTag = await runJS(`document.activeElement ? document.activeElement.tagName : ''`);
+  if (activeTag === 'IFRAME') {
+    return await _nativeTypeViaClipboard(text);
+  }
+
   // Use execCommand("insertText") — the ONLY approach that works for BOTH:
   // 1. Regular inputs/textareas (execCommand works natively)
   // 2. ContentEditable (ProseMirror/Draft.js/Slate) — execCommand causes real DOM mutation
@@ -1181,6 +1294,11 @@ export async function typeText({ text, selector, ref }) {
     `if('value' in el){var start=el.selectionStart||0;el.value=el.value.substring(0,start)+'${safeText}'+el.value.substring(el.selectionEnd||start);el.selectionStart=el.selectionEnd=start+${text.length};el.dispatchEvent(new Event('input',{bubbles:true}));el.dispatchEvent(new Event('change',{bubbles:true}));return 'Typed '+${text.length}+' chars via value set';}return 'Could not type';})()`
   );
 
+  // If JS typing failed and we're somehow in an iframe context, try native fallback
+  if (result === 'Could not type') {
+    return await _nativeTypeViaClipboard(text);
+  }
+
   return result;
 }