Brave Search: Privacy-First Web Search and Page Extraction Without a Browser

33,601 downloads, 136 stars. The Brave Search skill by @steipete is the top web search skill on ClawHub. It gives your AI agent access to Brave's independent search index plus URL content extraction — all via a two-command CLI with zero browser overhead.

The Problem It Solves

Web search in AI agents usually means one of two things: a full browser automation stack (heavy, slow, often blocked) or a search API wrapper that gives you titles and snippets but nothing useful about the actual page content.

Brave Search covers both needs with a minimal footprint. You get real search results from an independent index (not a Bing or Google reskin), and when you need the actual page content, content.js extracts it as clean markdown.

How It Works

The skill ships two Node.js scripts: search.js for search queries and content.js for URL extraction. Both run from the command line — no GUI, no browser process.

Setup:

cd ~/Projects/agent-scripts/skills/brave-search
npm ci
# Add to your .env or shell: BRAVE_API_KEY=your_key_here

Get your API key at brave.com/search/api. As of February 2026, Brave discontinued its free tier — new accounts start with $5/month in credits (~1,000 queries). Existing free-plan users were grandfathered in.

Core Commands

Search

./search.js "query"                  # Basic search, 5 results
./search.js "query" -n 10            # Up to 10 results
./search.js "query" --content        # Include full page content
./search.js "query" -n 3 --content   # 3 results with page content

Content Extraction

./content.js https://example.com/article

Fetches the URL and returns readable content as markdown — stripping navigation, ads, and boilerplate. Good for when a search result looks promising but you need the full text.

Output Format

--- Result 1 ---
Title: Page Title
Link: https://example.com/page
Snippet: Description from search results
Content: (if --content flag used)
  Markdown content extracted from the page...

--- Result 2 ---
...

Clean, parseable, no JSON wrestling.

Why Brave Search Specifically

Brave runs an independent web index — not a licensed reskin of Google or Bing. That matters for a few reasons:

• Different ranking signals — Brave deprioritizes SEO-optimized spam that game other indices
• Privacy posture — No user tracking by default; queries aren't logged to an ad profile
• Independent freshness — Brave's crawl schedule is separate from Google's, so you may find different freshness characteristics on niche topics

For factual lookups and documentation searches, the independence from Google's index is a practical benefit, not just a philosophical one.

Real-World Use Cases

Documentation lookup — Search for API references or error messages without leaving the terminal. The --content flag pulls full page content, not just snippets.

Fact verification — Cross-check claims against current web results before including them in generated content.

Competitive research — Search for a company or product and extract page content from their site for analysis.

Link checking — Use content.js to verify that URLs in your output still resolve and contain expected content.

Research pipelines — Chain search and extract: search for a topic, get URLs, extract each URL's content, summarize.

Comparison

Practical Tips

Use --content selectively. Content extraction adds time and API cost per result. Use it when you need the full article, not for initial scanning. A good pattern: search first (no --content), identify the relevant result, then extract just that URL with content.js.

Increase result count for broad topics. For research tasks where you're not sure which source is authoritative, use -n 10 and let your agent pick the best result.

Combine with file operations. Have your agent save extracted content to a file for longer research sessions:

./content.js https://docs.example.com/api > extracted.md

Check the API dashboard. The skill doesn't track your usage — check your Brave dashboard periodically to avoid hitting your monthly credit limit.

For local documentation — Use content.js even for pages you own. It's a clean way to extract structured content from any URL, including localhost if you're running a local server.

Considerations

• No free tier (as of Feb 2026) — Brave discontinued its free search API tier. All new accounts are on metered billing starting at $5/month (~1,000 queries). Factor this into production agent cost planning.
• Dynamic pages — content.js uses HTTP fetching, not a real browser. JavaScript-heavy single-page apps that render content client-side may return incomplete content.
• Search relevance — Brave's index is excellent for technical and English-language content but may be thinner on niche international topics compared to Google.
• BRAVE_API_KEY required — Unlike some skills that have fallback modes, this one requires an API key from day one.

The Bigger Picture

Web search is one of those capabilities where agents quickly outgrow toy implementations. Brave Search hits a practical sweet spot: real search quality, content extraction built in, independent index, and a lightweight CLI interface.

33,000+ downloads suggests it's become a default for agents that need to look things up — not just as a one-off, but as a regular part of workflows.

View the skill on ClawHub: brave-search