safari-mcp 2.8.9 → 2.9.1

package/extension/background.js

@@ -285,6 +285,19 @@ async function handleCommand(type, payload) {
       return targetTab.title;
     }
 
+    // --- Reload Extension (hot-reload after code changes) ---
+    // Allows the MCP server to trigger the extension to reload its own code from disk,
+    // bypassing the need for manual Safari → Preferences → Extensions → toggle.
+    // The WebSocket will disconnect as a side effect; the extension auto-reconnects.
+    case "reload_extension": {
+      // Respond BEFORE reload so the MCP server sees a success result.
+      // Delay the actual reload by a tick so the response can flush over the wire.
+      setTimeout(() => {
+        try { browser.runtime.reload(); } catch (_) { chrome.runtime.reload(); }
+      }, 50);
+      return { reloaded: true, version: browser.runtime.getManifest().version };
+    }
+
     case "read_page": {
       return await execInTab((sel, maxLen) => {
         if (sel) {
@@ -1287,7 +1300,53 @@ async function handleCommand(type, payload) {
           }
         }
         if (!root) return "Element not found: " + rootSelector;
-        let tree = walk(root, 0);
+
+        // TOP-LAYER MODAL DETECTION: when a modal/dialog is open, user intent is almost always
+        // to interact with it, not the page behind. Pages like Google Business Profile, Drive,
+        // Airtable rich dialogs open top-layer content that standard DOM walks miss or bury.
+        // Detect visible modals and, if found, walk them FIRST so their refs appear at the top.
+        let topLayerTree = "";
+        if (!rootSelector) {
+          const seenModals = new Set();
+          const modalSelectors = [
+            'dialog[open]',
+            '[role="dialog"]:not([aria-hidden="true"])',
+            '[role="alertdialog"]:not([aria-hidden="true"])',
+            '[aria-modal="true"]',
+            '[data-radix-dialog-content]',
+            '[data-headlessui-state*="open"][role="dialog"]',
+            '.MuiDialog-container',
+            // Google overlays: Search/GBP/Drive editors use these markers
+            '[jscontroller][aria-modal]',
+            '[jscontroller][role="dialog"]',
+            'c-wiz[role="dialog"]',
+            'c-wiz[aria-modal]'
+          ];
+          for (const sel of modalSelectors) {
+            let nodes;
+            try { nodes = document.querySelectorAll(sel); } catch (_) { continue; }
+            for (const m of nodes) {
+              if (seenModals.has(m)) continue;
+              seenModals.add(m);
+              const cs = window.getComputedStyle(m);
+              if (cs.display === "none" || cs.visibility === "hidden" || cs.opacity === "0") continue;
+              const r = m.getBoundingClientRect();
+              // A real modal covers a meaningful area
+              if (r.width < 150 || r.height < 100) continue;
+              // Tombstones: modals at 0,0 with 0 size are placeholders
+              if (r.width === 0 || r.height === 0) continue;
+              // Nested modals: skip if we already included a parent
+              let isNested = false;
+              for (const other of seenModals) {
+                if (other !== m && other.contains(m)) { isNested = true; break; }
+              }
+              if (isNested) continue;
+              topLayerTree += walk(m, 0);
+            }
+          }
+        }
+
+        let tree = topLayerTree + walk(root, 0);
         // Shadow roots are now walked INLINE inside walk() — no separate walkShadows needed.
         // Walk same-origin iframes
         const iframes = document.querySelectorAll("iframe");

package/extension/manifest.json

@@ -2,7 +2,7 @@
   "manifest_version": 3,
   "name": "Safari MCP Bridge",
   "description": "Connects AI agents to Safari — fast JS execution, screenshots, clicks with your real cookies/logins",
-  "version": "2.7.2",
+  "version": "2.9.1",
   "icons": {
     "48": "images/icon-48.png",
     "96": "images/icon-96.png",

package/index.js

@@ -666,6 +666,8 @@ const _nullMeansFailure = new Set([
 const _noOwnershipCheck = new Set([
   // Tab management
   "new_tab", "list_tabs", "close_tab", "switch_tab",
+  // Extension self-management (doesn't touch tabs)
+  "reload_extension",
   // Read-only — don't modify the page
   "read_page", "get_source", "snapshot", "accessibility_snapshot",
   "get_element", "query_all", "screenshot", "screenshot_element",
@@ -1256,6 +1258,19 @@ server.tool(
   }
 );
 
+server.tool(
+  "safari_reload_extension",
+  "Hot-reload the Safari MCP Bridge extension — forces it to reload its own code from disk without requiring manual Safari Preferences → Extensions → toggle. Use after editing extension/background.js or extension/content.js in the safari-mcp repo. The extension briefly disconnects during reload and auto-reconnects within ~2 seconds. NOTE: this tool itself requires the extension version already installed to support the `reload_extension` command (added in v2.9.1+). If your extension is older, trigger a manual reload once to pick up this feature.",
+  {},
+  async () => {
+    const result = await extensionOrFallback(
+      "reload_extension", {},
+      async () => "Extension fallback not available — this command requires the Safari MCP Bridge extension."
+    );
+    return { content: [{ type: "text", text: typeof result === 'string' ? result : JSON.stringify(result) }] };
+  }
+);
+
 server.tool(
   "safari_new_tab",
   "Open a new tab, optionally with a URL",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "safari-mcp",
-  "version": "2.8.9",
+  "version": "2.9.1",
   "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",