browse-agent-cli 0.0.3 → 0.0.5

package/README.md

@@ -36,6 +36,9 @@ browse-agent setup
 # Launch browser session
 browse-agent launch
 
+# Start only background service (do not launch browser)
+browse-agent launch --server-only
+
 # Navigate and extract
 browse-agent navigate "https://example.com"
 browse-agent get-content --format text
@@ -109,6 +112,7 @@ import {
 
 - `--browser <name>`: `chrome | chromium | edge | brave`
 - `--headless`
+- `--server-only`
 - `--port <number>`
 - `--servicePort <number>`
 - `--timeout <ms>`

package/dist/cli.js

@@ -77,6 +77,10 @@ function parseArgs(argv) {
 			flags.headless = true;
 			continue;
 		}
+		if (arg === "--server-only") {
+			flags["server-only"] = true;
+			continue;
+		}
 		if (arg === "--all") {
 			flags.all = true;
 			continue;
@@ -129,6 +133,7 @@ Feature commands (require launch first):
 Options:
   --browser <name>       Browser: chrome | chromium | edge | brave (default: chrome)
   --headless             Run in headless mode
+  --server-only          Start background service only, skip browser launch
   --port <number>        WebSocket port for browser-agent (default: 9315)
   --servicePort <number> Local service port (default: 9316)
   --tabId <id>           Target tab ID (from navigate or tabs list output)
@@ -142,6 +147,7 @@ Options:
 Examples:
   browse-agent setup
   browse-agent launch --browser edge --headless
+  browse-agent launch --server-only
   browse-agent navigate https://example.com
   browse-agent get-content --format text
   browse-agent get-dom "h1" --property innerText
@@ -299,6 +305,15 @@ try {
 		case "launch": {
 			const service = await ensureServiceRunning(flags);
 			const servicePort = service.servicePort;
+			if (flags["server-only"] === true) {
+				const status = await requestService(servicePort, "/status", "GET");
+				console.log(JSON.stringify({
+					servicePort,
+					mode: "server-only",
+					...status
+				}, null, 2));
+				break;
+			}
 			if (!service.newlyStarted) {
 				const status = await requestService(servicePort, "/status", "GET");
 				if (status.running === true) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "browse-agent-cli",
-  "version": "0.0.3",
+  "version": "0.0.5",
   "type": "module",
   "description": "TypeScript CLI for browse-agent",
   "main": "./dist/cli.js",
@@ -37,5 +37,15 @@
     "tsdown": "^0.13.3",
     "typescript": "^5.4.0"
   },
+  "bugs": {
+    "url": "https://github.com/imlinhanchao/browse-agent/issues"
+  },
+  "homepage": "https://github.com/imlinhanchao/browse-agent#readme",
+  "author": {
+    "name": "Hancel.Lin",
+    "email": "imlinhanchao@foxmail.com",
+    "url": "https://github.com/imlinhanchao"
+  },
+  "keywords": ["chrome", "browser", "automation", "websocket", "extension"],
   "license": "MIT"
 }

package/skills/SKILL.md

@@ -6,7 +6,7 @@ argument-hint: "URL to visit, or describe what data to extract from the web"
 
 # Browse Agent — Web Browsing & Data Extraction
 
-Control a real Chrome browser to visit web pages, extract content, take screenshots, query DOM elements, and run JavaScript — all via `browse-agent-sdk`.
+Control a real Chrome browser to visit web pages, extract content, take screenshots, query DOM elements, and run JavaScript.
 
 ## Prerequisites
 
@@ -19,7 +19,7 @@ Control a real Chrome browser to visit web pages, extract content, take screensh
 browse-agent setup
 ```
 
-Auto-detects local vs global mode. Use `--global` to force global installation in `~/.browse-agent/`. The setup checks existing state — if already installed, it exits immediately.
+Auto-detects local vs global mode. The setup checks existing state — if already installed, it exits immediately.
 
 ## Usage Procedure — Step-by-Step (Recommended)
 
@@ -28,7 +28,7 @@ Browse **interactively** — one command per step. Observe each result before de
 ### Step 1: Launch Browser
 
 ```bash
-browse-agent launch 2>/dev/null
+browse-agent launch
 ```
 
 Returns session info as JSON. The browser stays open until you close it.
@@ -39,22 +39,22 @@ Run each command separately. Read the output, then decide the next action:
 
 ```bash
 # Open a page — returns { tabId, url, title }
-browse-agent navigate "https://example.com" 2>/dev/null
+browse-agent navigate "https://example.com"
 
 # Read the page content — returns { content, url, title }
-browse-agent get-content --format text 2>/dev/null
+browse-agent get-content --format text
 
 # Specify a tab by ID (from navigate output) — use --tabId with any feature command
-browse-agent get-content --format text --tabId 123 2>/dev/null
+browse-agent get-content --format text --tabId 123
 
 # Query specific elements — returns { result }
-browse-agent get-dom "h1" --property innerText 2>/dev/null
+browse-agent get-dom "h1" --property innerText
 
 # Run JavaScript — returns { result }
-browse-agent evaluate "document.title" 2>/dev/null
+browse-agent evaluate "document.title"
 
 # Take a screenshot — returns { data (base64), format, width, height }
-browse-agent screenshot visible 2>/dev/null
+browse-agent screenshot visible
 ```
 
 **Key principle**: Each command outputs JSON to stdout. Use `--tabId <id>` to target a specific tab (ID comes from `navigate` or `tabs list` output). Parse the output, reason about it, then choose the next command. Don't pre-plan the entire interaction.
@@ -62,7 +62,7 @@ browse-agent screenshot visible 2>/dev/null
 ### Step 3: Close Browser
 
 ```bash
-browse-agent close 2>/dev/null
+browse-agent close
 ```
 
 ### Typical Interaction Flow
@@ -108,7 +108,7 @@ await browse(async (agent) => {
 });
 ```
 
-Run with: `node _browse_task.mjs 2>/dev/null`
+Run with: `node _browse_task.mjs`
 
 For full API and script examples, see [API Reference](./references/api.md) and [Examples](./references/examples.md).
 

package/skills/references/examples.md

@@ -6,79 +6,79 @@
 
 ```bash
 # 1. Launch browser
-browse-agent launch 2>/dev/null
+browse-agent launch
 
 # 2. Navigate to page
-browse-agent navigate "https://example.com" 2>/dev/null
+browse-agent navigate "https://example.com"
 
 # 3. Read content — decide next step based on output
-browse-agent get-content --format text 2>/dev/null
+browse-agent get-content --format text
 
 # 4. Done — close browser
-browse-agent close 2>/dev/null
+browse-agent close
 ```
 
 ### Find Specific Content on a Page
 
 ```bash
-browse-agent launch 2>/dev/null
-browse-agent navigate "https://news.ycombinator.com" 2>/dev/null
+browse-agent launch
+browse-agent navigate "https://news.ycombinator.com"
 
 # First, read the page to understand structure
-browse-agent get-content --format text 2>/dev/null
+browse-agent get-content --format text
 
 # Then query specific elements based on what you found
-browse-agent get-dom ".titleline > a" --property innerText --all 2>/dev/null
+browse-agent get-dom ".titleline > a" --property innerText --all
 
-browse-agent close 2>/dev/null
+browse-agent close
 ```
 
 ### Screenshot and Inspect
 
 ```bash
-browse-agent launch 2>/dev/null
-browse-agent navigate "https://example.com" 2>/dev/null
+browse-agent launch
+browse-agent navigate "https://example.com"
 
 # Take a screenshot to see the page visually
-browse-agent screenshot visible 2>/dev/null
+browse-agent screenshot visible
 
 # Run JS to count elements, check state, etc.
-browse-agent evaluate "document.querySelectorAll('a').length" 2>/dev/null
+browse-agent evaluate "document.querySelectorAll('a').length"
 
-browse-agent close 2>/dev/null
+browse-agent close
 ```
 
 ### Multi-Page Exploration
 
 ```bash
-browse-agent launch 2>/dev/null
+browse-agent launch
 
 # Visit first page
-browse-agent navigate "https://example.com" 2>/dev/null
-browse-agent get-content --format text 2>/dev/null
+browse-agent navigate "https://example.com"
+browse-agent get-content --format text
 
 # Visit second page (based on what you found)
-browse-agent navigate "https://example.org" 2>/dev/null
-browse-agent get-content --format text 2>/dev/null
+browse-agent navigate "https://example.org"
+browse-agent get-content --format text
 
 # Manage tabs
-browse-agent tabs list 2>/dev/null
-browse-agent tabs close 123 2>/dev/null
+browse-agent tabs list
+browse-agent tabs close 123
 
-browse-agent close 2>/dev/null
+browse-agent close
 ```
 
 ### Use Logged-in Browser Profile
 
 ```bash
 # Launch with user's default browser profile (preserves cookies/sessions)
-browse-agent launch --browser chrome 2>/dev/null
+browse-agent launch --browser chrome
 # ⚠ Close all Chrome windows first!
 
-USE_USER_PROFILE=true browse-agent launch 2>/dev/null
-browse-agent navigate "https://github.com/notifications" 2>/dev/null
-browse-agent get-content --format text 2>/dev/null
-browse-agent close 2>/dev/null
+USE_USER_PROFILE=true browse-agent launch
+browse-agent navigate "https://github.com/notifications"
+browse-agent get-content --format text
+browse-agent close
 ```
 
 ## One-Shot Script Examples
@@ -173,13 +173,13 @@ await browse(async (agent) => {
 
 ```bash
 # Use user's default Chrome profile (keeps login sessions)
-USE_USER_PROFILE=true node _browse_task.mjs 2>/dev/null
+USE_USER_PROFILE=true node _browse_task.mjs
 
 # Use Edge with user profile
-BROWSER=edge USE_USER_PROFILE=true node _browse_task.mjs 2>/dev/null
+BROWSER=edge USE_USER_PROFILE=true node _browse_task.mjs
 
 # Custom executable path
-CHROME_PATH=/path/to/browser node _browse_task.mjs 2>/dev/null
+CHROME_PATH=/path/to/browser node _browse_task.mjs
 ```
 
 ### Options via Code