@agent360/browser-mcp - npm Package Compare versions

Comparing version 1.16.1 to 1.19.0

extension/manifest.json

 {
   "manifest_version": 3,
   "name": "Agent360 Browser MCP",
-  "version": "1.16.0",
-  "description": "Control your real Chrome from Claude Code — navigate, click, fill, screenshot, solve CAPTCHAs. 24 tools, multi-session, human-in-the-loop.",
+  "version": "1.19.0",
+  "description": "Control your real Chrome from Claude Code — navigate, click, fill, set dates, dismiss overlays, autocomplete, upload, screenshot, solve CAPTCHAs. 33 tools, multi-session, human-in-the-loop.",
   "permissions": [
@@ -7,0 +7,0 @@ "tabs",

extension/offscreen.js

@@ -12,3 +12,3 @@ /**
 const BASE_PORT = 9876;
-const MAX_PORT = 9885;
+const MAX_PORT = 9895;
 const connections = new Map(); // port → WebSocket
@@ -96,4 +96,18 @@
 
+// Listen for terminate signals from background.js (sent when last tab in a session closes)
+chrome.runtime.onMessage.addListener((msg) => {
+  if (msg.type !== 'terminate_mcp_session' || typeof msg.port !== 'number') return;
+  const ws = connections.get(msg.port);
+  if (!ws) return;
+  try {
+    if (ws.readyState === WebSocket.OPEN) {
+      ws.send(JSON.stringify({ type: 'terminate' }));
+    }
+  } catch {}
+  try { ws.close(); } catch {}
+  // ws.onclose handler removes from connections + notifies background
+});
+
 // Initial scan + frequent rescan for new servers
 scanPorts();
 setInterval(scanPorts, 2000);

index.js

@@ -17,6 +17,12 @@ #!/usr/bin/env node
 import { execSync } from 'child_process';
-import { dirname } from 'path';
+import { dirname, join } from 'path';
 import { fileURLToPath } from 'url';
+import { readFileSync } from 'fs';
 import { TOOLS, PROVIDER_PAGES } from './tools.js';
 
+// Read version from package.json — single source of truth, never drifts
+const PKG_VERSION = JSON.parse(
+  readFileSync(join(dirname(fileURLToPath(import.meta.url)), 'package.json'), 'utf8')
+).version;
+
 // ── Auto-update on startup ─────────────────────────────────────────────────
@@ -51,3 +57,3 @@
 const BASE_PORT = 9876;
-const MAX_PORT = 9885;
+const MAX_PORT = 9895; // 20 ports instead of 10 — zombies die within 5s via parent check
 let extensionSocket = null;
@@ -95,2 +101,8 @@ let activePort = null;
       try { msg = JSON.parse(data.toString()); } catch { return; }
+
+      if (msg.type === 'terminate') {
+        process.stderr.write('[MCP] Terminate signal received from extension (last tab closed) — exiting\n');
+        process.exit(0);
+      }
+
       const { id, result, error } = msg;
@@ -134,8 +146,12 @@ const p = pending.get(id);
 
-function sendToExtension(method, params = {}, timeoutMs = 30000) {
+async function sendToExtension(method, params = {}, timeoutMs = 30000, _retries = 5) {
+  // Retry if extension is temporarily disconnected (reconnects every 2s)
+  if (!extensionSocket || extensionSocket.readyState !== 1) {
+    if (_retries > 0) {
+      await new Promise(r => setTimeout(r, 1500));
+      return sendToExtension(method, params, timeoutMs, _retries - 1);
+    }
+    throw new Error('Chrome extension not connected after 5 retries. Open Chrome and ensure Agent360 Browser MCP extension is installed.');
+  }
   return new Promise((resolve, reject) => {
-    if (!extensionSocket || extensionSocket.readyState !== 1) {
-      reject(new Error('Chrome extension not connected. Open Chrome and ensure Agent360 Browser MCP extension is installed.'));
-      return;
-    }
     const id = ++cmdId;
@@ -213,2 +229,8 @@ const timer = setTimeout(() => {
 
+## Hard inputs — use the specialised tools first
+- **Date inputs** → use browser_set_date (NOT browser_fill). Handles native date inputs, masked text inputs (MM/DD/YYYY etc.), AND calendar pickers (MUI, react-datepicker, AntD, Lexical/Meta). 3-path fallback with read-back verification.
+- **Autocomplete / combobox** (Languages on Meta Ads, country selects, async dropdowns) → use browser_set_combobox (NOT browser_select_option). Types partial query, waits for filtered listbox, clicks option. Supports multi-value chips.
+- **Drag-drop file zones without visible file input** → use browser_drop_file (NOT browser_upload_file). Finds hidden input in subtree/parent.
+- **Annoying popups blocking the flow** (cookie banners, "Don't show again", Advantage+ tooltips, draft-confirm prompts) → call browser_dismiss_overlays before each major step. It only clicks safe close affordances by default; preserves forms with editable text fields.
+
 ## When things fail
@@ -219,2 +241,3 @@ - Element not found → try text-based selector instead of CSS
 - CAPTCHA blocks page → use browser_ask_user, let human solve it
+- browser_fill seemingly succeeds but value reverts → switch to browser_set_date or browser_set_combobox (most reverts are React-controlled validators)
 
@@ -228,3 +251,3 @@ ## Extension updates
 const mcpServer = new Server(
-  { name: 'agent360-browser', version: '1.16.0' },
+  { name: 'agent360-browser', version: PKG_VERSION },
   { capabilities: { tools: {} } },
@@ -272,2 +295,6 @@ { instructions: INSTRUCTIONS },
       browser_solve_captcha: 'solve_captcha',
+      browser_set_date: 'set_date',
+      browser_dismiss_overlays: 'dismiss_overlays',
+      browser_set_combobox: 'set_combobox',
+      browser_drop_file: 'drop_file',
     };
@@ -356,3 +383,16 @@
 
-// Detect Claude Code exit (stdin closes when conversation ends)
+// Detect Claude Code exit — check if parent process is still alive
+// stdin.on('end') doesn't work because MCP SDK's StdioServerTransport owns stdin
+const parentPid = process.ppid;
+const parentCheck = setInterval(() => {
+  try {
+    process.kill(parentPid, 0); // signal 0 = check if process exists
+  } catch {
+    process.stderr.write(`[MCP] Parent process ${parentPid} died — shutting down\n`);
+    clearInterval(parentCheck);
+    process.exit(0);
+  }
+}, 5000); // check every 5 seconds
+
+// Also listen for stdin close as backup
 process.stdin.on('end', () => {
@@ -359,0 +399,0 @@ process.stderr.write('[MCP] stdin closed — shutting down\n');

package.json

 {
   "name": "@agent360/browser-mcp",
-  "version": "1.16.1",
+  "version": "1.19.0",
   "description": "Browser MCP — control your real Chrome from Claude Code. 29 tools, CAPTCHA solving, file upload, multi-session, human-in-the-loop.",
@@ -5,0 +5,0 @@ "mcpName": "io.github.Agent360dk/browser-mcp",

tools.js

@@ -59,3 +59,3 @@ /**
     name: 'browser_fill',
-    description: 'Fill a form input field with a value. Supports CSS selectors AND text-based selectors. Auto-scrolls and focuses the element. Works on CSP-strict sites via Chrome Debugger API.',
+    description: 'Fill a form input field with a value. Supports CSS selectors AND text-based selectors. Auto-scrolls and focuses the element. Works on CSP-strict sites via Chrome Debugger API. For date inputs use browser_set_date, for autocomplete/combobox use browser_set_combobox.',
     inputSchema: {
@@ -124,3 +124,3 @@ type: 'object',
     name: 'browser_select_option',
-    description: 'Select an option from a dropdown menu. Works with native <select> elements AND custom dropdowns (Angular Material, React Select, etc.). For custom dropdowns: clicks the trigger, waits for options, then clicks the matching option by text.',
+    description: 'Select an option from a dropdown menu. Works with native <select> elements AND custom dropdowns (Angular Material, React Select, etc.). For custom dropdowns: clicks the trigger, waits for options, then clicks the matching option by text. For autocomplete (typing filters options) use browser_set_combobox instead.',
     inputSchema: {
@@ -137,2 +137,55 @@ type: 'object',
   {
+    name: 'browser_dismiss_overlays',
+    description: 'Dismiss visible popups, modals, tooltips, banners, and "Are you sure?"-style overlays in one call. Heuristic-based: finds close affordance via aria-label, text content (Skip/Cancel/Ikke nu/Don\'t show/Got it/Close), or × character button. Use when a flow is interrupted by unexpected dialogs (cookie banners, onboarding tooltips, draft-confirm prompts on Meta Ads, etc.). Returns list of what was dismissed.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        scope: { type: 'string', enum: ['non_critical', 'aggressive'], description: 'non_critical (default): skip dialogs containing editable form inputs (preserves user data). aggressive: dismiss everything.' },
+        max_passes: { type: 'number', description: 'Number of dismissal passes (some overlays reveal others when closed). Default: 3' },
+      },
+    },
+  },
+  {
+    name: 'browser_set_combobox',
+    description: 'Set value(s) on an autocomplete/combobox input. Handles the click → type query → wait for filtered listbox → click option flow as one MCP call. Supports multi-select (e.g., Languages on Meta Ads). Use when browser_select_option fails because options render lazily after typing.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        selector: { type: 'string', description: 'CSS selector for the combobox/autocomplete input' },
+        value: { type: 'string', description: 'Single value to select (use this OR values)' },
+        values: { type: 'array', items: { type: 'string' }, description: 'Array of values for multi-select. E.g. ["Danish", "English", "Swedish"]' },
+        multi: { type: 'boolean', description: 'True if combobox accepts multiple values (chips). Default: auto-detected from presence of values array' },
+        query_chars: { type: 'number', description: 'How many characters to type as filter query (default: 4 or full value length, whichever is smaller)' },
+        wait_ms: { type: 'number', description: 'Max ms to wait for options listbox to appear after typing (default: 3000)' },
+      },
+      required: ['selector'],
+    },
+  },
+  {
+    name: 'browser_drop_file',
+    description: 'Upload a file by finding a hidden <input type="file"> within a drag-drop zone\'s subtree (or parent up to 2 levels). Use when browser_upload_file fails because the dropzone has no visible file input. Returns clear error if no input is found anywhere — pure drop-zones without backing inputs require manual handling.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        selector: { type: 'string', description: 'CSS selector for the drop-zone target element (e.g. ".upload-area")' },
+        file: { type: 'string', description: 'Single absolute file path' },
+        files: { type: 'array', items: { type: 'string' }, description: 'Array of absolute file paths' },
+      },
+      required: ['selector'],
+    },
+  },
+  {
+    name: 'browser_set_date',
+    description: 'Robustly set a date input — handles native <input type="date">, masked text inputs (e.g. MM/DD/YYYY), and calendar pickers (MUI, react-datepicker, AntD, Lexical/Meta). Tries native value-set, format-aware typing via Input.insertText, and ARIA-based picker navigation in sequence with read-back verification. Use instead of browser_fill when fill fails or for any input that opens a calendar widget.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        selector: { type: 'string', description: 'CSS selector for the date input element' },
+        date: { type: 'string', description: 'ISO date string (YYYY-MM-DD), e.g. "2026-05-15"' },
+        skip_picker: { type: 'boolean', description: 'If true, only try native + masked paths and skip calendar-picker navigation (default: false)' },
+      },
+      required: ['selector', 'date'],
+    },
+  },
+  {
     name: 'browser_handle_dialog',
@@ -313,3 +366,3 @@ description: 'Handle JavaScript alert(), confirm(), or prompt() dialogs. Call this BEFORE triggering the action that causes the dialog. Waits for the dialog to appear, then accepts or dismisses it.',
     name: 'browser_upload_file',
-    description: 'Upload a file to a <input type="file"> element on the page. Uses Chrome Debugger API to set files programmatically — no dialog needed.',
+    description: 'Upload a file to a <input type="file"> element on the page. Uses Chrome Debugger API to set files programmatically — no dialog needed. For drag-drop zones without visible file input use browser_drop_file.',
     inputSchema: {
@@ -316,0 +369,0 @@ type: 'object',

New alerts

Filesystem access

Supply chain risk

Accesses the file system, and could potentially read sensitive data.

AI-detected potential code anomaly

Supply chain risk

AI has identified unusual behaviors that may pose a security risk.

Found 2 instances

Long strings

Supply chain risk

Contains long string literals, which may be a sign of obfuscated or packed code.

URL strings

Supply chain risk

Package contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.

Fixed alerts

AI-detected potential malware

Supply chain risk

AI has identified this package as malware. This is a strong signal that the package may be malicious.

AI-detected potential code anomaly

Supply chain risk

AI has identified unusual behaviors that may pose a security risk.

Long strings

Supply chain risk

Contains long string literals, which may be a sign of obfuscated or packed code.

URL strings

Supply chain risk

Package contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.

Improved metrics

Total package byte prevSize

186228

32.38%

Lines of code

3515

38.77%

Number of high supply chain risk alerts

0

-100%

Worsened metrics

Number of low supply chain risk alerts

12

33.33%

No dependency changes