Power Search

Self-hosted research tool combining Brave Search API + Browserless content fetching. Search the web with optional full-page content extraction and HTML parsing.

Self-hosted web research tool combining Brave Search API and Browserless for content fetching. Search quickly, fetch full content optionally, extract structured data from HTML.

Features

• Fast search: Brave Search API (10 results in <1s)
• Content fetching: Browserless headless browser for full-page extraction
• HTML parsing: Extract clean text from web pages
• CLI interface: Simple command-line tool
• Telegram integration: Ready for OpenClaw routing
• Offline-capable: Local Browserless, no cloud dependencies

Installation

clawhub install power-search

Or manually:

git clone https://github.com/yourusername/power-search.git
cd power-search
npm install

Setup

1. Docker & Browserless

# Install Docker
sudo apt-get install docker.io docker-compose

# Start Browserless container
docker run -d --name browserless --restart unless-stopped \
  -p 3000:3000 browserless/chrome:latest

2. Brave Search API Key

Get a free key from search.brave.com:

• 100 free queries/day
• Set environment variable: export BRAVE_API_KEY=your_key_here

3. Configuration (Optional)

Create .env in the skill directory:

BRAVE_API_KEY=your_key_here
BROWSERLESS_HOST=http://localhost
BROWSERLESS_PORT=3000

Usage

Search Only

search "nodejs frameworks"
search "machine learning" --limit 20

Search + Fetch Content

search "nodejs express" --fetch
search "rust web frameworks" --fetch --limit 5

Verbose Mode

search "web scraping" --fetch --verbose

Options

• --fetch — Fetch full content from top results
• --limit N — Number of results (default: 10, max: 20)
• --timeout MS — Timeout in milliseconds (default: 10000)
• --verbose — Show detailed logging

Output

Search Results

=== Search Results for "nodejs frameworks" ===

[1] The 5 Most Popular Node.js Web Frameworks in 2025 - DEV Community
    https://dev.to/leapcell/the-5-most-popular-nodejs-web-frameworks-in-2025-12po
    Nest.js is a framework known for building scalable and efficient Node.js...

[2] Express - Node.js web application framework
    https://expressjs.com/
    Fast, unopinionated, minimalist web framework for Node.js...

...

With Content Fetch

📄 Content from "nodejs express"

[1] Express - Node.js web application framework
    URL: https://expressjs.com/
    Content:
    Express - Node.js web application framework Getting started...

API (Programmatic Usage)

const BraveSearch = require('./scripts/brave-search');
const Browserless = require('./scripts/browserless');

const brave = new BraveSearch('your_api_key');
const results = await brave.search('nodejs', 10);

const browserless = new Browserless('http://localhost', 3000);
const content = await browserless.fetchContent('https://example.com');

Telegram Integration

Power Search is OpenClaw-ready. Once registered:

/search artemis 2
/search machine learning --fetch --limit 5

Results post directly to Telegram with formatted output.

Architecture

search/
├── scripts/
│   ├── search.js              # Main CLI entry point
│   ├── search-runner.js       # JSON output runner
│   ├── telegram-handler.js    # Telegram command handler
│   ├── brave-search.js        # Brave API wrapper
│   └── browserless.js         # Browserless wrapper
├── package.json
├── SKILL.md                   # This file
└── openclaw.json              # OpenClaw manifest

Performance

Operation	Time	Notes
Search	<1s	Brave API is very fast
Fetch	2-10s	Depends on page complexity
Extract	<500ms	HTML parsing + text cleanup

Limitations

• Some sites block headless browsers (Cloudflare, etc.)
• JavaScript-heavy sites may not render full content
• Brave Search free tier: 100 queries/day
• HTML extraction is basic (good for text, may miss complex layouts)

Troubleshooting

Browserless not running

docker ps | grep browserless
docker start browserless
docker logs browserless

Rate limit exceeded

Wait 24 hours or upgrade to Brave's paid plan.

Slow results

• Reduce --limit to fetch fewer results
• Increase timeout: --timeout 30000
• Check Docker memory: docker stats browserless

Future Enhancements

• AI summaries (Claude/GPT integration)
• Research chains (auto follow-ups)
• Result caching
• Advanced exports (PDF, HTML with styling)
• Saved searches & monitoring

License

MIT

Author

Kaito - Digital Assistant