💬 Commit message: Update 2026-02-06 20:28:30, 13 files, 1401 lines

📁 Files changed: 13
📝 Lines changed: 1401

  • .gitignore
  • CLAUDE.md
  • PLAN.md
  • README.md
  • TODO.md
  • package-lock.json
  • package.json
  • browser.ts
  • cli.ts
  • index.ts
  • server.ts
  • types.ts
  • tsconfig.json
This commit is contained in:
Adam Ladachowski
2026-02-06 20:28:30 +01:00
commit eccc03ee7a
13 changed files with 1401 additions and 0 deletions
+4
View File
@@ -0,0 +1,4 @@
/node_modules/
/public/
/dist/
/screenshots/
+44
View File
@@ -0,0 +1,44 @@
# CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
## Build & Development
```bash
npm install # Install dependencies
npx playwright install webkit # Install WebKit browser
npm run build # Compile TypeScript to dist/
npm run dev # Watch mode compilation
npm start # Run the CLI (after build)
```
## Architecture
This is a TypeScript library providing headless browser automation via Playwright WebKit. It has three modes of operation:
1. **CLI** (`src/cli.ts`) - Direct command-line usage for screenshots, clicking, typing, and querying
2. **Server** (`src/server.ts`) - HTTP server accepting JSON commands via POST requests
3. **Library** (`src/index.ts`) - Programmatic API via `ClaudeBrowser` class
### Core Components
- **ClaudeBrowser** (`src/browser.ts`) - Main class wrapping Playwright WebKit. Handles browser lifecycle, navigation, and all DOM interactions. Uses `executeCommand()` to process typed command objects.
- **BrowserServer** (`src/server.ts`) - HTTP server that wraps ClaudeBrowser. Accepts JSON POST requests and returns JSON responses. CORS-enabled.
- **Types** (`src/types.ts`) - Discriminated union of command types (`BrowserCommand`) and response types (`CommandResponse`).
### Command Flow
Commands are typed objects with a `cmd` discriminator:
```typescript
{ cmd: 'goto', url: string }
{ cmd: 'click', selector: string }
{ cmd: 'type', selector: string, text: string }
{ cmd: 'query', selector: string }
// etc.
```
All commands return `{ ok: true, ...data }` or `{ ok: false, error: string }`.
## CLI Options
Key flags: `-s <port>` (server mode), `-q <selector>` (query), `-c <selector>` (click), `-t <sel>=<text>` (type), `-i` (interactive), `--headed` (visible browser).
+238
View File
@@ -0,0 +1,238 @@
# Plan: Web Debugging Commands
## Phase 1: Storage & Session Commands
### Description
Add commands for inspecting and manipulating browser storage and cookies. These are essential for debugging authentication, session state, and client-side data persistence.
### Steps
#### Step 1.1: Add Cookies Command
- **Objective**: Implement get/set/clear cookies functionality
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `CookiesCommand` type with `action: 'get' | 'set' | 'clear'`
- Add optional `name`, `value`, `url` fields for set operation
- Implement `cookies()` method using `context.cookies()` and `context.addCookies()`
- Add logging with cookie icon
#### Step 1.2: Add Storage Command
- **Objective**: Implement localStorage and sessionStorage read/write/clear
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: Step 1.1
- **Implementation**:
- Add `StorageCommand` type with `storage: 'local' | 'session'` and `action: 'get' | 'set' | 'clear'`
- Implement via `page.evaluate()` accessing `window.localStorage` / `window.sessionStorage`
- Return all items on get, or specific key if provided
## Phase 2: Console & Debugging Commands
### Description
Add commands for capturing console output and debugging page state. Critical for understanding client-side errors and application behavior.
### Steps
#### Step 2.1: Add Console Command
- **Objective**: Capture and return console messages (log, error, warn, info)
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `ConsoleCommand` type with optional `clear: boolean`
- Store console messages in array via `page.on('console')`
- Initialize listener in `launch()` method
- Return accumulated messages with type, text, and timestamp
#### Step 2.2: Add Metrics Command
- **Objective**: Return performance metrics and DOM statistics
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `MetricsCommand` type
- Use `page.evaluate()` to gather `performance.timing`, DOM node counts
- Return structured metrics object
## Phase 3: Interaction Commands
### Description
Add commands for advanced page interactions beyond click and type. Enables testing hover states, keyboard shortcuts, and form controls.
### Steps
#### Step 3.1: Add Scroll Command
- **Objective**: Scroll to position, element, or by delta
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `ScrollCommand` with `selector?: string`, `x?: number`, `y?: number`, `behavior?: 'smooth' | 'instant'`
- If selector provided, use `element.scrollIntoView()`
- Otherwise use `window.scrollTo()`
#### Step 3.2: Add Hover Command
- **Objective**: Trigger hover state on element
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `HoverCommand` with `selector: string`
- Use `page.hover(selector)`
- Return success with element info
#### Step 3.3: Add Select Command
- **Objective**: Select option(s) in dropdown/select elements
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `SelectCommand` with `selector: string`, `value: string | string[]`
- Use `page.selectOption(selector, value)`
- Return selected values
#### Step 3.4: Add Keys Command
- **Objective**: Send keyboard shortcuts and special keys
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `KeysCommand` with `keys: string` (e.g., "Control+a", "Escape", "Enter")
- Use `page.keyboard.press()` for single keys
- Support key combinations
## Phase 4: Network Commands
### Description
Add commands for network inspection and manipulation. Enables debugging API calls, blocking resources, and simulating network conditions.
### Steps
#### Step 4.1: Add Network Logging
- **Objective**: Capture and return network requests/responses
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `NetworkCommand` with `action: 'log' | 'clear'`, optional `filter: string`
- Store requests via `page.on('request')` and `page.on('response')`
- Return array of {url, method, status, type, timing}
#### Step 4.2: Add Block Command
- **Objective**: Block URLs or patterns (ads, analytics, etc.)
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: Step 4.1
- **Implementation**:
- Add `BlockCommand` with `patterns: string[]` and `action: 'add' | 'remove' | 'clear'`
- Use `page.route()` to abort matching requests
- Maintain list of blocked patterns
#### Step 4.3: Add Throttle Command
- **Objective**: Simulate slow network conditions
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `ThrottleCommand` with `preset: 'slow3g' | 'fast3g' | 'offline' | 'none'`
- Use CDP session to set network conditions via `Network.emulateNetworkConditions`
## Phase 5: Output Commands
### Description
Add commands for exporting page content in different formats.
### Steps
#### Step 5.1: Add PDF Command
- **Objective**: Save page as PDF
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `PdfCommand` with `path: string`, optional `format`, `landscape`, `margin`
- Use `page.pdf()` (WebKit supports this)
- Return path to saved file
#### Step 5.2: Add Accessibility Command
- **Objective**: Dump accessibility tree snapshot
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `AccessibilityCommand` with optional `selector` for subtree
- Use `page.accessibility.snapshot()`
- Return tree structure
## Phase 6: Advanced Commands
### Description
Add commands for viewport manipulation, device emulation, and frame handling.
### Steps
#### Step 6.1: Add Viewport Command
- **Objective**: Resize viewport dynamically
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `ViewportCommand` with `width: number`, `height: number`
- Use `page.setViewportSize()`
- Return new dimensions
#### Step 6.2: Add Emulate Command
- **Objective**: Emulate mobile devices
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: Step 6.1
- **Implementation**:
- Add `EmulateCommand` with `device: string` (e.g., 'iPhone 12', 'Pixel 5')
- Use Playwright's device descriptors
- Apply viewport, userAgent, deviceScaleFactor
#### Step 6.3: Add Frames Command
- **Objective**: List and switch to iframe contexts
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `FramesCommand` with `action: 'list' | 'switch'`, optional `selector` or `index`
- Use `page.frames()` to list, `frame.locator()` to switch context
- Track current frame for subsequent commands
#### Step 6.4: Add Dialog Command
- **Objective**: Handle alert/confirm/prompt dialogs
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `DialogCommand` with `action: 'accept' | 'dismiss'`, optional `text` for prompt
- Set up `page.on('dialog')` handler
- Queue dialogs and respond based on command
#### Step 6.5: Add Upload Command
- **Objective**: Fill file input elements
- **Files**: `src/types.ts`, `src/browser.ts`, `src/server.ts`, `src/index.ts`
- **Dependencies**: None
- **Implementation**:
- Add `UploadCommand` with `selector: string`, `files: string[]`
- Use `page.setInputFiles(selector, files)`
- Return success with file count
## Phase 7: Documentation & Polish
### Description
Update documentation and CLI help text with all new commands.
### Steps
#### Step 7.1: Update README
- **Objective**: Document all new commands with examples
- **Files**: `README.md`
- **Dependencies**: Phases 1-6
- **Implementation**:
- Add command reference section
- Include curl examples for each command
- Document response formats
#### Step 7.2: Update CLAUDE.md
- **Objective**: Update developer documentation
- **Files**: `CLAUDE.md`
- **Dependencies**: Step 7.1
- **Implementation**:
- Update architecture section with new command flow
- Document internal logging format
#### Step 7.3: Update CLI Help
- **Objective**: Update --help output with command list
- **Files**: `src/cli.ts`
- **Dependencies**: Phases 1-6
- **Implementation**:
- Expand server mode help section
- Group commands by category
+136
View File
@@ -0,0 +1,136 @@
# @aladac/claude-browse
Headless browser automation for Claude Code using Playwright WebKit.
## Installation
```bash
npm install @aladac/claude-browse
npx playwright install webkit
```
## CLI Usage
```bash
# Take a screenshot
claude-browse https://example.com
# Custom viewport and output
claude-browse -o page.png -w 1920 -h 1080 https://example.com
# Query elements
claude-browse -q "a[href]" https://example.com
claude-browse -q "img" -j https://example.com # JSON output
# Click and interact
claude-browse -c "button.submit" https://example.com
claude-browse -t "input[name=q]=hello" -c "button[type=submit]" https://google.com
# Chain actions
claude-browse -c ".cookie-accept" -c "a.nav-link" -q "h1" https://example.com
# Interactive mode (visible browser)
claude-browse -i --headed https://example.com
```
## Server Mode
Start a persistent browser server that accepts commands via HTTP:
```bash
claude-browse -s 3000 # headless
claude-browse -s 3000 --headed # visible browser
```
Send commands:
```bash
# Navigate
curl -X POST localhost:3000 -d '{"cmd":"goto","url":"https://example.com"}'
# Click
curl -X POST localhost:3000 -d '{"cmd":"click","selector":"button.submit"}'
# Type
curl -X POST localhost:3000 -d '{"cmd":"type","selector":"#search","text":"hello"}'
# Query elements
curl -X POST localhost:3000 -d '{"cmd":"query","selector":"a[href]"}'
# Screenshot
curl -X POST localhost:3000 -d '{"cmd":"screenshot","path":"page.png"}'
# Get current URL
curl -X POST localhost:3000 -d '{"cmd":"url"}'
# Get page HTML
curl -X POST localhost:3000 -d '{"cmd":"html"}'
# Navigation
curl -X POST localhost:3000 -d '{"cmd":"back"}'
curl -X POST localhost:3000 -d '{"cmd":"forward"}'
curl -X POST localhost:3000 -d '{"cmd":"reload"}'
# Wait
curl -X POST localhost:3000 -d '{"cmd":"wait","ms":2000}'
# Close server
curl -X POST localhost:3000 -d '{"cmd":"close"}'
```
## Programmatic Usage
```typescript
import { ClaudeBrowser, startServer } from '@aladac/claude-browse';
// Direct browser control
const browser = new ClaudeBrowser({
headless: true,
width: 1280,
height: 800,
});
await browser.launch();
await browser.goto('https://example.com');
const elements = await browser.query('a[href]');
console.log(elements);
await browser.click('button.submit');
await browser.type('#input', 'hello');
await browser.screenshot('page.png');
await browser.close();
// Or start a server
const server = await startServer({ port: 3000, headless: false });
// Server runs until closed via HTTP or SIGINT
```
## API
### ClaudeBrowser
- `launch()` - Launch the browser
- `close()` - Close the browser
- `goto(url)` - Navigate to URL
- `click(selector)` - Click element
- `type(selector, text)` - Type into input
- `query(selector)` - Query elements, returns attributes
- `screenshot(path?, fullPage?)` - Take screenshot
- `getUrl()` - Get current URL and title
- `getHtml(full?)` - Get page HTML
- `back()` / `forward()` / `reload()` - Navigation
- `wait(ms)` - Wait for timeout
- `newPage()` - Open new page
- `executeCommand(cmd)` - Execute a command object
### BrowserServer
- `start()` - Start HTTP server
- `stop()` - Stop server and close browser
- `getPort()` - Get server port
## License
MIT
+36
View File
@@ -0,0 +1,36 @@
# TODO: Web Debugging Commands
## Phase 1: Storage & Session Commands
- [ ] Step 1.1: Add Cookies Command
- [ ] Step 1.2: Add Storage Command
## Phase 2: Console & Debugging Commands
- [ ] Step 2.1: Add Console Command
- [ ] Step 2.2: Add Metrics Command
## Phase 3: Interaction Commands
- [ ] Step 3.1: Add Scroll Command
- [ ] Step 3.2: Add Hover Command
- [ ] Step 3.3: Add Select Command
- [ ] Step 3.4: Add Keys Command
## Phase 4: Network Commands
- [ ] Step 4.1: Add Network Logging
- [ ] Step 4.2: Add Block Command
- [ ] Step 4.3: Add Throttle Command
## Phase 5: Output Commands
- [ ] Step 5.1: Add PDF Command
- [ ] Step 5.2: Add Accessibility Command
## Phase 6: Advanced Commands
- [ ] Step 6.1: Add Viewport Command
- [ ] Step 6.2: Add Emulate Command
- [ ] Step 6.3: Add Frames Command
- [ ] Step 6.4: Add Dialog Command
- [ ] Step 6.5: Add Upload Command
## Phase 7: Documentation & Polish
- [ ] Step 7.1: Update README
- [ ] Step 7.2: Update CLAUDE.md
- [ ] Step 7.3: Update CLI Help
+101
View File
@@ -0,0 +1,101 @@
{
"name": "@aladac/claude-browse",
"version": "0.1.0",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "@aladac/claude-browse",
"version": "0.1.0",
"license": "MIT",
"dependencies": {
"playwright": "^1.41.0"
},
"bin": {
"claude-browse": "dist/cli.js"
},
"devDependencies": {
"@types/node": "^20.11.0",
"typescript": "^5.3.0"
},
"engines": {
"node": ">=18"
}
},
"node_modules/@types/node": {
"version": "20.19.32",
"resolved": "https://registry.npmjs.org/@types/node/-/node-20.19.32.tgz",
"integrity": "sha512-Ez8QE4DMfhjjTsES9K2dwfV258qBui7qxUsoaixZDiTzbde4U12e1pXGNu/ECsUIOi5/zoCxAQxIhQnaUQ2VvA==",
"dev": true,
"license": "MIT",
"dependencies": {
"undici-types": "~6.21.0"
}
},
"node_modules/fsevents": {
"version": "2.3.2",
"resolved": "https://registry.npmjs.org/fsevents/-/fsevents-2.3.2.tgz",
"integrity": "sha512-xiqMQR4xAeHTuB9uWm+fFRcIOgKBMiOBP+eXiyT7jsgVCq1bkVygt00oASowB7EdtpOHaaPgKt812P9ab+DDKA==",
"hasInstallScript": true,
"license": "MIT",
"optional": true,
"os": [
"darwin"
],
"engines": {
"node": "^8.16.0 || ^10.6.0 || >=11.0.0"
}
},
"node_modules/playwright": {
"version": "1.58.2",
"resolved": "https://registry.npmjs.org/playwright/-/playwright-1.58.2.tgz",
"integrity": "sha512-vA30H8Nvkq/cPBnNw4Q8TWz1EJyqgpuinBcHET0YVJVFldr8JDNiU9LaWAE1KqSkRYazuaBhTpB5ZzShOezQ6A==",
"license": "Apache-2.0",
"dependencies": {
"playwright-core": "1.58.2"
},
"bin": {
"playwright": "cli.js"
},
"engines": {
"node": ">=18"
},
"optionalDependencies": {
"fsevents": "2.3.2"
}
},
"node_modules/playwright-core": {
"version": "1.58.2",
"resolved": "https://registry.npmjs.org/playwright-core/-/playwright-core-1.58.2.tgz",
"integrity": "sha512-yZkEtftgwS8CsfYo7nm0KE8jsvm6i/PTgVtB8DL726wNf6H2IMsDuxCpJj59KDaxCtSnrWan2AeDqM7JBaultg==",
"license": "Apache-2.0",
"bin": {
"playwright-core": "cli.js"
},
"engines": {
"node": ">=18"
}
},
"node_modules/typescript": {
"version": "5.9.3",
"resolved": "https://registry.npmjs.org/typescript/-/typescript-5.9.3.tgz",
"integrity": "sha512-jl1vZzPDinLr9eUt3J/t7V6FgNEw9QjvBPdysz9KfQDD41fQrC2Y4vKQdiaUpFT4bXlb1RHhLpp8wtm6M5TgSw==",
"dev": true,
"license": "Apache-2.0",
"bin": {
"tsc": "bin/tsc",
"tsserver": "bin/tsserver"
},
"engines": {
"node": ">=14.17"
}
},
"node_modules/undici-types": {
"version": "6.21.0",
"resolved": "https://registry.npmjs.org/undici-types/-/undici-types-6.21.0.tgz",
"integrity": "sha512-iwDZqg0QAGrg9Rav5H4n0M64c3mkR59cJ6wQp+7C4nI0gsmExaedaYLNO44eT4AtBBwjbTiGPMlt2Md0T9H9JQ==",
"dev": true,
"license": "MIT"
}
}
}
+45
View File
@@ -0,0 +1,45 @@
{
"name": "@aladac/claude-browse",
"version": "0.1.0",
"description": "Headless browser automation for Claude Code using Playwright WebKit",
"type": "module",
"main": "dist/index.js",
"types": "dist/index.d.ts",
"bin": {
"claude-browse": "dist/cli.js"
},
"scripts": {
"build": "tsc",
"dev": "tsc --watch",
"start": "node dist/cli.js",
"prepublishOnly": "npm run build"
},
"keywords": [
"playwright",
"webkit",
"browser",
"automation",
"claude",
"headless"
],
"author": "aladac",
"license": "MIT",
"repository": {
"type": "git",
"url": "https://github.com/aladac/claude-browse"
},
"engines": {
"node": ">=18"
},
"dependencies": {
"playwright": "^1.41.0"
},
"devDependencies": {
"@types/node": "^20.11.0",
"typescript": "^5.3.0"
},
"files": [
"dist",
"README.md"
]
}
+208
View File
@@ -0,0 +1,208 @@
import { webkit, Browser, BrowserContext, Page } from 'playwright';
import { resolve } from 'node:path';
import type {
BrowserOptions,
BrowserCommand,
CommandResponse,
ElementInfo,
} from './types.js';
export class ClaudeBrowser {
private browser: Browser | null = null;
private context: BrowserContext | null = null;
private page: Page | null = null;
private options: Required<BrowserOptions>;
constructor(options: BrowserOptions = {}) {
this.options = {
headless: options.headless ?? true,
width: options.width ?? 1280,
height: options.height ?? 800,
};
}
async launch(): Promise<void> {
this.browser = await webkit.launch({ headless: this.options.headless });
this.context = await this.browser.newContext({
viewport: {
width: this.options.width,
height: this.options.height,
},
});
this.page = await this.context.newPage();
}
async close(): Promise<void> {
if (this.browser) {
await this.browser.close();
this.browser = null;
this.context = null;
this.page = null;
}
}
private ensurePage(): Page {
if (!this.page) {
throw new Error('Browser not launched. Call launch() first.');
}
return this.page;
}
async goto(url: string): Promise<{ url: string; title: string }> {
const page = this.ensurePage();
await page.goto(url, { waitUntil: 'networkidle' });
return { url: page.url(), title: await page.title() };
}
async click(selector: string): Promise<{ url: string }> {
const page = this.ensurePage();
await page.click(selector);
await page.waitForLoadState('networkidle').catch(() => {});
return { url: page.url() };
}
async type(selector: string, text: string): Promise<void> {
const page = this.ensurePage();
await page.fill(selector, text);
}
async query(selector: string): Promise<ElementInfo[]> {
const page = this.ensurePage();
return page.$$eval(selector, (nodes) =>
nodes.map((el) => {
const attrs: Record<string, string> = {};
for (const attr of el.attributes) {
attrs[attr.name] = attr.value;
}
return {
tag: el.tagName.toLowerCase(),
text: el.textContent?.trim().slice(0, 200) || '',
attributes: attrs,
};
})
);
}
async screenshot(
path?: string,
fullPage = false
): Promise<{ path: string; buffer?: Buffer }> {
const page = this.ensurePage();
const resolvedPath = resolve(path || 'screenshot.png');
const buffer = await page.screenshot({ path: resolvedPath, fullPage });
return { path: resolvedPath, buffer };
}
async getUrl(): Promise<{ url: string; title: string }> {
const page = this.ensurePage();
return { url: page.url(), title: await page.title() };
}
async getHtml(full = false): Promise<string> {
const page = this.ensurePage();
const html = await page.content();
return full ? html : html.slice(0, 10000);
}
async back(): Promise<{ url: string }> {
const page = this.ensurePage();
await page.goBack();
return { url: page.url() };
}
async forward(): Promise<{ url: string }> {
const page = this.ensurePage();
await page.goForward();
return { url: page.url() };
}
async reload(): Promise<{ url: string }> {
const page = this.ensurePage();
await page.reload();
return { url: page.url() };
}
async wait(ms = 1000): Promise<void> {
const page = this.ensurePage();
await page.waitForTimeout(ms);
}
async newPage(): Promise<void> {
if (!this.context) {
throw new Error('Browser not launched. Call launch() first.');
}
this.page = await this.context.newPage();
}
async eval(script: string): Promise<unknown> {
const page = this.ensurePage();
return page.evaluate(script);
}
async executeCommand(cmd: BrowserCommand): Promise<CommandResponse> {
try {
switch (cmd.cmd) {
case 'goto': {
const result = await this.goto(cmd.url);
return { ok: true, ...result };
}
case 'click': {
const result = await this.click(cmd.selector);
return { ok: true, ...result };
}
case 'type': {
await this.type(cmd.selector, cmd.text);
return { ok: true };
}
case 'query': {
const elements = await this.query(cmd.selector);
return { ok: true, count: elements.length, elements };
}
case 'screenshot': {
const result = await this.screenshot(cmd.path, cmd.fullPage);
return { ok: true, path: result.path };
}
case 'url': {
const result = await this.getUrl();
return { ok: true, ...result };
}
case 'html': {
const html = await this.getHtml(cmd.full);
return { ok: true, html };
}
case 'back': {
const result = await this.back();
return { ok: true, ...result };
}
case 'forward': {
const result = await this.forward();
return { ok: true, ...result };
}
case 'reload': {
const result = await this.reload();
return { ok: true, ...result };
}
case 'wait': {
await this.wait(cmd.ms);
return { ok: true };
}
case 'newpage': {
await this.newPage();
return { ok: true };
}
case 'close': {
await this.close();
return { ok: true };
}
case 'eval': {
const result = await this.eval(cmd.script);
return { ok: true, result };
}
default:
return { ok: false, error: `Unknown command: ${(cmd as any).cmd}` };
}
} catch (err) {
return { ok: false, error: (err as Error).message };
}
}
}
+199
View File
@@ -0,0 +1,199 @@
#!/usr/bin/env node
import { parseArgs } from 'node:util';
import { resolve } from 'node:path';
import { ClaudeBrowser } from './browser.js';
import { startServer } from './server.js';
const { values, positionals } = parseArgs({
allowPositionals: true,
options: {
output: { type: 'string', short: 'o', default: 'screenshot.png' },
width: { type: 'string', short: 'w', default: '1280' },
height: { type: 'string', short: 'h', default: '800' },
fullpage: { type: 'boolean', short: 'f', default: false },
wait: { type: 'string', default: '2000' },
headed: { type: 'boolean', default: false },
interactive: { type: 'boolean', short: 'i', default: false },
query: { type: 'string', short: 'q' },
json: { type: 'boolean', short: 'j', default: false },
click: { type: 'string', short: 'c', multiple: true },
type: { type: 'string', short: 't', multiple: true },
serve: { type: 'string', short: 's' },
help: { type: 'boolean', default: false },
version: { type: 'boolean', short: 'v', default: false },
},
});
const HELP = `
Usage: claude-browse [options] <url>
Options:
-o, --output <file> Output screenshot path (default: screenshot.png)
-w, --width <px> Viewport width (default: 1280)
-h, --height <px> Viewport height (default: 800)
-f, --fullpage Capture full page scroll
--wait <ms> Wait time after load (default: 2000)
--headed Show browser window
-i, --interactive Keep browser open for manual interaction
-q, --query <selector> Query elements by CSS selector and show attributes
-j, --json Output query results as JSON
-c, --click <selector> Click on element (can be repeated for multiple clicks)
-t, --type <sel>=<text> Type text into input (can be repeated)
-s, --serve <port> Start browser server on port (default: 3000)
-v, --version Show version
--help Show this help
Examples:
claude-browse https://example.com
claude-browse -o page.png -w 1920 -h 1080 https://example.com
claude-browse -i --headed https://example.com
claude-browse -q "a[href]" https://example.com
claude-browse -q "img" -j https://example.com
claude-browse -c "button.submit" https://example.com
claude-browse -t "input[name=q]=hello" -c "button[type=submit]" https://google.com
claude-browse -c ".cookie-accept" -c "a.nav-link" -q "h1" https://example.com
Server mode:
claude-browse -s 3000 # Start server on port 3000
claude-browse -s 3000 --headed # Start with visible browser
# Send commands via curl:
curl -X POST http://localhost:3000 -d '{"cmd":"goto","url":"https://example.com"}'
curl -X POST http://localhost:3000 -d '{"cmd":"click","selector":"button"}'
curl -X POST http://localhost:3000 -d '{"cmd":"type","selector":"input","text":"hello"}'
curl -X POST http://localhost:3000 -d '{"cmd":"query","selector":"a[href]"}'
curl -X POST http://localhost:3000 -d '{"cmd":"screenshot","path":"shot.png"}'
curl -X POST http://localhost:3000 -d '{"cmd":"url"}'
curl -X POST http://localhost:3000 -d '{"cmd":"html"}'
curl -X POST http://localhost:3000 -d '{"cmd":"close"}'
`;
async function main(): Promise<void> {
if (values.version) {
console.log('claude-browse 0.1.0');
process.exit(0);
}
// Server mode
if (values.serve) {
const port = parseInt(values.serve) || 3000;
const server = await startServer({
port,
headless: !values.headed,
width: parseInt(values.width as string),
height: parseInt(values.height as string),
});
process.on('SIGINT', async () => {
console.log('\nShutting down...');
await server.stop();
process.exit(0);
});
// Keep alive
await new Promise(() => {});
return;
}
if (values.help || positionals.length === 0) {
console.log(HELP);
process.exit(0);
}
const url = positionals[0];
const outputPath = resolve(values.output as string);
const browser = new ClaudeBrowser({
headless: !values.headed,
width: parseInt(values.width as string),
height: parseInt(values.height as string),
});
await browser.launch();
console.log(`Navigating to: ${url}`);
await browser.goto(url);
await browser.wait(parseInt(values.wait as string));
// Process type actions (before clicks, typically for form filling)
const typeActions = values.type as string[] | undefined;
if (typeActions?.length) {
for (const typeAction of typeActions) {
const eqIndex = typeAction.indexOf('=');
if (eqIndex === -1) {
console.error(
`Invalid --type format: "${typeAction}" (expected selector=text)`
);
continue;
}
const selector = typeAction.slice(0, eqIndex);
const text = typeAction.slice(eqIndex + 1);
console.log(`Typing "${text}" into: ${selector}`);
await browser.type(selector, text);
}
}
// Process click actions
const clickActions = values.click as string[] | undefined;
if (clickActions?.length) {
for (const selector of clickActions) {
console.log(`Clicking: ${selector}`);
await browser.click(selector);
await browser.wait(500);
}
const { url: currentUrl } = await browser.getUrl();
console.log(`Current URL: ${currentUrl}`);
}
// Query elements if -q/--query is specified
if (values.query) {
const elements = await browser.query(values.query);
if (values.json) {
console.log(JSON.stringify(elements, null, 2));
} else {
console.log(
`Found ${elements.length} element(s) matching "${values.query}":\n`
);
elements.forEach((el, i) => {
console.log(`[${i + 1}] <${el.tag}>`);
for (const [name, value] of Object.entries(el.attributes)) {
console.log(` ${name}="${value}"`);
}
if (el.text) {
console.log(
` text: "${el.text.slice(0, 100)}${el.text.length > 100 ? '...' : ''}"`
);
}
console.log();
});
}
await browser.close();
process.exit(0);
}
if (!values.interactive) {
console.log(`Saving screenshot to: ${outputPath}`);
await browser.screenshot(outputPath, values.fullpage);
await browser.close();
console.log('Done!');
} else {
console.log('Interactive mode - browser will stay open.');
console.log('Press Ctrl+C to exit.');
process.on('SIGINT', async () => {
console.log('\nClosing browser...');
await browser.close();
process.exit(0);
});
// Keep alive
await new Promise(() => {});
}
}
main().catch((err) => {
console.error('Error:', err.message);
process.exit(1);
});
+25
View File
@@ -0,0 +1,25 @@
export { ClaudeBrowser } from './browser.js';
export { BrowserServer, startServer } from './server.js';
export type {
BrowserOptions,
BrowserCommand,
CommandResponse,
ElementInfo,
SuccessResponse,
ErrorResponse,
GotoCommand,
ClickCommand,
TypeCommand,
QueryCommand,
ScreenshotCommand,
UrlCommand,
HtmlCommand,
BackCommand,
ForwardCommand,
ReloadCommand,
WaitCommand,
NewPageCommand,
CloseCommand,
EvalCommand,
} from './types.js';
export type { ServerOptions } from './server.js';
+233
View File
@@ -0,0 +1,233 @@
import { createServer, Server, IncomingMessage, ServerResponse } from 'node:http';
import { ClaudeBrowser } from './browser.js';
import type { BrowserCommand, BrowserOptions, CommandResponse } from './types.js';
export interface ServerOptions extends BrowserOptions {
port?: number;
}
// ANSI colors
const c = {
reset: '\x1b[0m',
bold: '\x1b[1m',
dim: '\x1b[2m',
cyan: '\x1b[36m',
green: '\x1b[32m',
yellow: '\x1b[33m',
blue: '\x1b[34m',
magenta: '\x1b[35m',
red: '\x1b[31m',
gray: '\x1b[90m',
white: '\x1b[37m',
bgBlue: '\x1b[44m',
bgGreen: '\x1b[42m',
bgYellow: '\x1b[43m',
bgMagenta: '\x1b[45m',
};
function timestamp(): string {
return c.gray + '[' + new Date().toISOString() + ']' + c.reset;
}
function logCommand(cmd: BrowserCommand): void {
const ts = timestamp();
switch (cmd.cmd) {
case 'goto':
console.log(`${ts} ${c.cyan}${c.bold}${c.reset} ${c.cyan}GOTO${c.reset} ${c.white}${cmd.url}${c.reset}`);
break;
case 'click':
console.log(`${ts} ${c.yellow}${c.bold}${c.reset} ${c.yellow}CLICK${c.reset} ${c.white}${cmd.selector}${c.reset}`);
break;
case 'type':
console.log(`${ts} ${c.magenta}${c.bold}${c.reset} ${c.magenta}TYPE${c.reset} ${c.white}${cmd.selector}${c.reset} ${c.dim}="${cmd.text}"${c.reset}`);
break;
case 'query':
console.log(`${ts} ${c.blue}${c.bold}?${c.reset} ${c.blue}QUERY${c.reset} ${c.white}${cmd.selector}${c.reset}`);
break;
case 'screenshot':
console.log(`${ts} ${c.green}${c.bold}📷${c.reset} ${c.green}SCREENSHOT${c.reset} ${c.dim}${cmd.path || 'screenshot.png'}${c.reset}`);
break;
case 'url':
console.log(`${ts} ${c.cyan}${c.bold}🔗${c.reset} ${c.cyan}URL${c.reset}`);
break;
case 'html':
console.log(`${ts} ${c.blue}${c.bold}<>${c.reset} ${c.blue}HTML${c.reset} ${cmd.full ? c.dim + '(full)' + c.reset : ''}`);
break;
case 'back':
console.log(`${ts} ${c.yellow}${c.bold}${c.reset} ${c.yellow}BACK${c.reset}`);
break;
case 'forward':
console.log(`${ts} ${c.yellow}${c.bold}${c.reset} ${c.yellow}FORWARD${c.reset}`);
break;
case 'reload':
console.log(`${ts} ${c.yellow}${c.bold}${c.reset} ${c.yellow}RELOAD${c.reset}`);
break;
case 'wait':
console.log(`${ts} ${c.gray}${c.bold}${c.reset} ${c.gray}WAIT${c.reset} ${c.dim}${cmd.ms || 1000}ms${c.reset}`);
break;
case 'newpage':
console.log(`${ts} ${c.green}${c.bold}+${c.reset} ${c.green}NEW PAGE${c.reset}`);
break;
case 'close':
console.log(`${ts} ${c.red}${c.bold}${c.reset} ${c.red}CLOSE${c.reset}`);
break;
case 'eval':
const preview = cmd.script.length > 50 ? cmd.script.slice(0, 50) + '...' : cmd.script;
console.log(`${ts} ${c.magenta}${c.bold}${c.reset} ${c.magenta}EVAL${c.reset} ${c.dim}${preview}${c.reset}`);
break;
}
}
function logResult(cmd: BrowserCommand, result: CommandResponse): void {
const ts = timestamp();
if (!result.ok) {
console.log(`${ts} ${c.red}✗ Error: ${result.error}${c.reset}`);
return;
}
switch (cmd.cmd) {
case 'goto':
if ('title' in result && result.title) {
console.log(`${ts} ${c.green}${c.reset} ${c.dim}${result.title}${c.reset}`);
}
break;
case 'click':
if ('url' in result) {
console.log(`${ts} ${c.green}${c.reset} ${c.dim}${result.url}${c.reset}`);
}
break;
case 'query':
if ('count' in result) {
console.log(`${ts} ${c.green}${c.reset} ${c.dim}Found ${result.count} element(s)${c.reset}`);
}
break;
case 'screenshot':
if ('path' in result) {
console.log(`${ts} ${c.green}${c.reset} ${c.dim}Saved to ${result.path}${c.reset}`);
}
break;
case 'url':
if ('url' in result) {
console.log(`${ts} ${c.green}${c.reset} ${c.dim}${result.url}${c.reset}`);
}
break;
case 'html':
if ('html' in result) {
console.log(`${ts} ${c.green}${c.reset} ${c.dim}${result.html?.length || 0} chars${c.reset}`);
}
break;
case 'eval':
if ('result' in result) {
const json = JSON.stringify(result.result);
const preview = json && json.length > 80 ? json.slice(0, 80) + '...' : json;
console.log(`${ts} ${c.green}${c.reset} ${c.dim}${preview}${c.reset}`);
}
break;
default:
console.log(`${ts} ${c.green}${c.reset}`);
}
}
export class BrowserServer {
private browser: ClaudeBrowser;
private server: Server | null = null;
private port: number;
constructor(options: ServerOptions = {}) {
this.browser = new ClaudeBrowser(options);
this.port = options.port ?? 3000;
}
async start(): Promise<void> {
await this.browser.launch();
this.server = createServer(
async (req: IncomingMessage, res: ServerResponse) => {
// CORS headers
res.setHeader('Access-Control-Allow-Origin', '*');
res.setHeader('Access-Control-Allow-Methods', 'POST, OPTIONS');
res.setHeader('Access-Control-Allow-Headers', 'Content-Type');
if (req.method === 'OPTIONS') {
res.writeHead(204);
res.end();
return;
}
if (req.method !== 'POST') {
res.writeHead(405, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ ok: false, error: 'POST only' }));
return;
}
let body = '';
for await (const chunk of req) {
body += chunk;
}
try {
const cmd: BrowserCommand = JSON.parse(body);
logCommand(cmd);
// Handle close specially to shut down server
if (cmd.cmd === 'close') {
const result = { ok: true as const };
logResult(cmd, result);
res.writeHead(200, { 'Content-Type': 'application/json' });
res.end(JSON.stringify(result));
await this.stop();
process.exit(0);
}
const result = await this.browser.executeCommand(cmd);
logResult(cmd, result);
res.writeHead(200, { 'Content-Type': 'application/json' });
res.end(JSON.stringify(result));
} catch (err) {
const errorResult = { ok: false as const, error: (err as Error).message };
console.log(`${timestamp()} ${c.red}${errorResult.error}${c.reset}`);
res.writeHead(500, { 'Content-Type': 'application/json' });
res.end(JSON.stringify(errorResult));
}
}
);
return new Promise((resolve) => {
this.server!.listen(this.port, () => {
console.log();
console.log(`${c.bold}${c.cyan} 🌐 Claude Browse Server${c.reset}`);
console.log(`${c.dim} ─────────────────────────${c.reset}`);
console.log(` ${c.green}${c.reset} Listening on ${c.bold}http://localhost:${this.port}${c.reset}`);
console.log();
console.log(`${c.dim} Commands:${c.reset}`);
console.log(` ${c.cyan}goto${c.reset} ${c.yellow}click${c.reset} ${c.magenta}type${c.reset} ${c.blue}query${c.reset} ${c.green}screenshot${c.reset}`);
console.log(` ${c.cyan}url${c.reset} ${c.blue}html${c.reset} ${c.yellow}back${c.reset} ${c.yellow}forward${c.reset} ${c.yellow}reload${c.reset} ${c.gray}wait${c.reset} ${c.red}close${c.reset}`);
console.log();
console.log(`${c.dim} Example:${c.reset}`);
console.log(` ${c.gray}curl -X POST localhost:${this.port} -d '{"cmd":"goto","url":"https://example.com"}'${c.reset}`);
console.log();
resolve();
});
});
}
async stop(): Promise<void> {
console.log(`\n${c.dim} Shutting down...${c.reset}`);
if (this.server) {
this.server.close();
this.server = null;
}
await this.browser.close();
console.log(` ${c.green}${c.reset} Browser closed\n`);
}
getPort(): number {
return this.port;
}
}
export async function startServer(options: ServerOptions = {}): Promise<BrowserServer> {
const server = new BrowserServer(options);
await server.start();
return server;
}
+113
View File
@@ -0,0 +1,113 @@
export interface BrowserOptions {
headless?: boolean;
width?: number;
height?: number;
}
export interface ElementInfo {
tag: string;
text: string;
attributes: Record<string, string>;
}
// Command types
export interface GotoCommand {
cmd: 'goto';
url: string;
}
export interface ClickCommand {
cmd: 'click';
selector: string;
}
export interface TypeCommand {
cmd: 'type';
selector: string;
text: string;
}
export interface QueryCommand {
cmd: 'query';
selector: string;
}
export interface ScreenshotCommand {
cmd: 'screenshot';
path?: string;
fullPage?: boolean;
}
export interface UrlCommand {
cmd: 'url';
}
export interface HtmlCommand {
cmd: 'html';
full?: boolean;
}
export interface BackCommand {
cmd: 'back';
}
export interface ForwardCommand {
cmd: 'forward';
}
export interface ReloadCommand {
cmd: 'reload';
}
export interface WaitCommand {
cmd: 'wait';
ms?: number;
}
export interface NewPageCommand {
cmd: 'newpage';
}
export interface CloseCommand {
cmd: 'close';
}
export interface EvalCommand {
cmd: 'eval';
script: string;
}
export type BrowserCommand =
| GotoCommand
| ClickCommand
| TypeCommand
| QueryCommand
| ScreenshotCommand
| UrlCommand
| HtmlCommand
| BackCommand
| ForwardCommand
| ReloadCommand
| WaitCommand
| NewPageCommand
| CloseCommand
| EvalCommand;
// Response types
export interface SuccessResponse {
ok: true;
url?: string;
title?: string;
path?: string;
html?: string;
count?: number;
elements?: ElementInfo[];
result?: unknown;
}
export interface ErrorResponse {
ok: false;
error: string;
}
export type CommandResponse = SuccessResponse | ErrorResponse;
+19
View File
@@ -0,0 +1,19 @@
{
"compilerOptions": {
"target": "ES2022",
"module": "NodeNext",
"moduleResolution": "NodeNext",
"lib": ["ES2022"],
"outDir": "dist",
"rootDir": "src",
"strict": true,
"esModuleInterop": true,
"skipLibCheck": true,
"forceConsistentCasingInFileNames": true,
"declaration": true,
"declarationMap": true,
"sourceMap": true
},
"include": ["src/**/*"],
"exclude": ["node_modules", "dist"]
}