abu-browser-bridge 0.5.2 → 0.6.6

package/dist/index.js

@@ -17,6 +17,7 @@ import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
 import { startWSServer, stopWSServer } from './wsServer.js';
 import { registerTools } from './tools.js';
+import { PKG_VERSION } from './version.js';
 const DEFAULT_WS_PORT = 9876;
 const DISCOVERY_PORT = 9875;
 /**
@@ -123,7 +124,7 @@ async function main() {
     // 2. Create MCP server
     const mcpServer = new McpServer({
         name: 'abu-browser-bridge',
-        version: '0.5.2',
+        version: PKG_VERSION,
     });
     // 3. Register browser tools
     registerTools(mcpServer);

package/dist/tools.js

@@ -67,7 +67,7 @@ export function registerTools(server) {
     });
     // 2. browser_snapshot
     server.tool('snapshot', `Get a structured snapshot of all interactive elements on the page (buttons, inputs, links, selects, etc.). Returns each element with a short reference ID (e.g., "e1") that can be used in subsequent actions. This is the primary way to understand what's on a page before taking action.`, {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         selector: z.string().optional().describe('Optional CSS selector to scope the snapshot to a specific area of the page'),
     }, async ({ tabId, selector }) => {
         ensureConnected();
@@ -76,7 +76,7 @@ export function registerTools(server) {
     });
     // 3. browser_click
     server.tool('click', 'Click an element on the page. Returns the result of the click action.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         locator: z.string().describe(`JSON string of element locator. ${LocatorDescription}`),
     }, async ({ tabId, locator }) => {
         ensureConnected();
@@ -86,7 +86,7 @@ export function registerTools(server) {
     });
     // 4. browser_fill
     server.tool('fill', 'Fill in a text input, textarea, or other editable field. Clears existing content and types the new value, triggering proper input/change events for framework compatibility (React, Vue, etc.).', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         locator: z.string().describe(`JSON string of element locator. ${LocatorDescription}`),
         value: z.string().describe('The text value to fill into the field'),
     }, async ({ tabId, locator, value }) => {
@@ -97,7 +97,7 @@ export function registerTools(server) {
     });
     // 5. browser_select
     server.tool('select', 'Select an option from a <select> dropdown element.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         locator: z.string().describe(`JSON string of element locator. ${LocatorDescription}`),
         value: z.string().describe('The option value or visible text to select'),
     }, async ({ tabId, locator, value }) => {
@@ -108,14 +108,14 @@ export function registerTools(server) {
     });
     // 6. browser_wait_for
     server.tool('wait_for', `Wait for a condition to be met on the page. Useful for waiting for elements to appear after a click, waiting for loading to complete, or waiting for page navigation. Returns when the condition is met or times out.`, {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         condition: z.string().describe(`JSON string of wait condition. Options:
 - { "type": "appear", "locator": { "text": "成功" } } — wait for element to appear
 - { "type": "disappear", "locator": { "css": ".loading" } } — wait for element to disappear
 - { "type": "enabled", "locator": { "text": "提交" } } — wait for element to become clickable
 - { "type": "textContains", "locator": { "css": "#status" }, "text": "完成" } — wait for text content
 - { "type": "urlContains", "pattern": "/success" } — wait for URL change`),
-        timeout: z.number().optional().default(30000).describe('Maximum wait time in ms (default: 30000)'),
+        timeout: z.coerce.number().optional().default(30000).describe('Maximum wait time in ms (default: 30000)'),
     }, async ({ tabId, condition, timeout }) => {
         ensureConnected();
         const parsed = parseCondition(condition);
@@ -124,7 +124,7 @@ export function registerTools(server) {
     });
     // 7. browser_extract_text
     server.tool('extract_text', 'Extract text content from the page or a specific element. Useful for reading content, checking values, or verifying results.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         selector: z.string().optional().describe('CSS selector to extract text from. If omitted, extracts the full page text (may be large).'),
     }, async ({ tabId, selector }) => {
         ensureConnected();
@@ -133,7 +133,7 @@ export function registerTools(server) {
     });
     // 8. browser_extract_table
     server.tool('extract_table', 'Extract structured data from an HTML table on the page. Returns headers and rows as arrays.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         selector: z.string().optional().describe('CSS selector for the target table. If omitted, extracts the largest table on the page.'),
     }, async ({ tabId, selector }) => {
         ensureConnected();
@@ -142,9 +142,9 @@ export function registerTools(server) {
     });
     // 9. browser_scroll
     server.tool('scroll', 'Scroll the page or a specific element.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         direction: z.enum(['up', 'down', 'left', 'right']).describe('Scroll direction'),
-        amount: z.number().optional().default(500).describe('Scroll amount in pixels (default: 500)'),
+        amount: z.coerce.number().optional().default(500).describe('Scroll amount in pixels (default: 500)'),
         selector: z.string().optional().describe('CSS selector for the scrollable element. If omitted, scrolls the whole page.'),
     }, async ({ tabId, direction, amount, selector }) => {
         ensureConnected();
@@ -153,7 +153,7 @@ export function registerTools(server) {
     });
     // 10. browser_navigate
     server.tool('navigate', 'Navigate a tab to a specific URL, or go back/forward in history.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         url: z.string().optional().describe('URL to navigate to. Omit for back/forward.'),
         action: z.enum(['goto', 'back', 'forward', 'reload']).optional().default('goto').describe('Navigation action (default: goto)'),
     }, async ({ tabId, url, action }) => {
@@ -163,7 +163,7 @@ export function registerTools(server) {
     });
     // 11. browser_keyboard
     server.tool('keyboard', 'Send keyboard events to the page. Supports key combinations.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         key: z.string().describe('Key to press (e.g., "Enter", "Tab", "Escape", "a", "ArrowDown")'),
         modifiers: z.array(z.enum(['ctrl', 'shift', 'alt', 'meta'])).optional().describe('Modifier keys to hold'),
     }, async ({ tabId, key, modifiers }) => {
@@ -173,7 +173,7 @@ export function registerTools(server) {
     });
     // 12. browser_execute_js
     server.tool('execute_js', 'Execute arbitrary JavaScript code in the context of the page. Use this as a fallback when other tools cannot achieve the desired result. Returns the result of the expression.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
         code: z.string().describe('JavaScript code to execute. The last expression value is returned.'),
     }, async ({ tabId, code }) => {
         ensureConnected();
@@ -182,7 +182,7 @@ export function registerTools(server) {
     });
     // 13. browser_screenshot
     server.tool('screenshot', 'Take a screenshot of the visible area of a tab. Returns a base64-encoded PNG image. Useful for visual confirmation of actions.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
     }, async ({ tabId }) => {
         ensureConnected();
         const res = await sendToExtension('screenshot', { tabId });
@@ -197,7 +197,25 @@ export function registerTools(server) {
         }
         return { content: [{ type: 'text', text: formatResult(res) }] };
     });
-    // 14. browser_connection_status
+    // 14. browser_screenshot_full_page
+    server.tool('screenshot_full_page', 'Take a full-page screenshot by scrolling and stitching the entire page content. Returns a base64-encoded PNG image of the complete page. Use this when the user asks for a "long screenshot" or wants to capture content beyond the visible viewport. This is slower than a regular screenshot.', {
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
+    }, async ({ tabId }) => {
+        ensureConnected();
+        // Full-page capture needs more time: scroll + multiple captures + stitch
+        const res = await sendToExtension('screenshot_full_page', { tabId }, 120_000);
+        if (res.success && typeof res.data === 'string') {
+            return {
+                content: [{
+                        type: 'image',
+                        data: res.data.replace(/^data:image\/png;base64,/, ''),
+                        mimeType: 'image/png',
+                    }]
+            };
+        }
+        return { content: [{ type: 'text', text: formatResult(res) }] };
+    });
+    // 15. browser_connection_status
     server.tool('connection_status', 'Check whether the Chrome Extension is connected to this bridge. Use this to verify the extension is ready before performing browser actions.', async () => {
         const connected = isExtensionConnected();
         return {
@@ -217,7 +235,7 @@ export function registerTools(server) {
     });
     // 16. start_recording — record user interactions
     server.tool('start_recording', 'Start recording user interactions on a page (clicks, inputs, selects). The user performs actions manually, then call stop_recording to get a list of recorded steps that can be used as an automation template.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
     }, async ({ tabId }) => {
         ensureConnected();
         const res = await sendToExtension('start_recording', { tabId });
@@ -225,7 +243,7 @@ export function registerTools(server) {
     });
     // 17. stop_recording — stop recording and return captured steps
     server.tool('stop_recording', 'Stop recording user interactions and return the captured steps. Each step includes the action type, element locator, and value. Use these steps as a template to replay the automation.', {
-        tabId: z.number().describe('Tab ID from get_tabs'),
+        tabId: z.coerce.number().describe('Tab ID from get_tabs'),
     }, async ({ tabId }) => {
         ensureConnected();
         const res = await sendToExtension('stop_recording', { tabId });

package/dist/version.d.ts

@@ -0,0 +1 @@
+export declare const PKG_VERSION: string;

package/dist/version.js

@@ -0,0 +1,7 @@
+import { readFileSync } from 'fs';
+import { fileURLToPath } from 'url';
+import { dirname, resolve } from 'path';
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = dirname(__filename);
+const pkg = JSON.parse(readFileSync(resolve(__dirname, '../package.json'), 'utf-8'));
+export const PKG_VERSION = pkg.version;

package/dist/wsServer.js

@@ -12,10 +12,10 @@
 import { WebSocketServer, WebSocket } from 'ws';
 import { createServer } from 'http';
 import { randomBytes } from 'crypto';
+import { PKG_VERSION } from './version.js';
 const DEFAULT_WS_PORT = 9876;
 const DISCOVERY_PORT = 9875;
 const HEARTBEAT_INTERVAL = 15_000; // 15s
-const PONG_TIMEOUT = 5_000;
 let wss = null;
 let discoveryServer = null;
 let extensionSocket = null;
@@ -70,7 +70,7 @@ function startDiscoveryServer() {
                     pid: process.pid,
                     extensionConnected: isExtensionConnected(),
                     uptime: Math.round((Date.now() - startTime) / 1000),
-                    version: '0.5.2',
+                    version: PKG_VERSION,
                     token: authToken,
                 }));
                 return;

package/package.json

@@ -1,13 +1,13 @@
 {
   "name": "abu-browser-bridge",
-  "version": "0.5.2",
+  "version": "0.6.6",
   "description": "MCP Server that bridges Abu AI assistant with Chrome Extension for browser automation",
   "type": "module",
-  "license": "MIT",
+  "license": "SEE LICENSE IN LICENSE",
   "author": "pm-shawn",
   "repository": {
     "type": "git",
-    "url": "https://github.com/anthropics/abu"
+    "url": "https://github.com/PM-Shawn/Abu-Cowork"
   },
   "keywords": ["mcp", "browser", "automation", "chrome-extension", "abu"],
   "bin": {