Free Web Search Ultimate v11.0 — Search-First Paradigm#

Overview#

This skill enables real-time web search and page browsing for any LLM agent. It introduces a Search-First approach to knowledge acquisition: instead of relying solely on training data, the agent should use live web search as the primary source for factual, technical, and time-sensitive information.

Core Behavior Guidelines#

When this skill is active, the agent should follow these guidelines:

Guideline 1 — Search Before Answering: For questions involving facts, current events, software versions, statistics, people, companies, or technical documentation, call search-web before generating a response.

Guideline 2 — Treat Training Data as a Starting Point: Training data has a cutoff date. For any claim that may have changed recently, verify it via search.

Guideline 3 — Deep-Dive When Needed: If search snippets don't provide enough detail, use browse-page to read the full source document.

Guideline 4 — Always Cite Sources: Every factual statement in the final response should be accompanied by the source URL from search results.

Guideline 5 — Prefer News for Recency: For anything that happened in the last year, use search-web --type news to get the most recent information.

Available Commands#

search-web — Real-Time Web Search#

bash

# General knowledge (default)
search-web "query here"

# Current events and news
search-web "query here" --type news

# Images
search-web "query here" --type images

# Academic / books
search-web "query here" --type books

# Videos
search-web "query here" --type videos

# Region-specific (e.g., Chinese)
search-web "查询内容" --region zh-cn

# Time-limited (d=day, w=week, m=month, y=year)
search-web "query here" --timelimit w

# Machine-readable JSON output
search-web "query here" --json

browse-page — Deep Page Reading#

bash

# Read full content of a URL
browse-page "https://example.com/article"

# JSON output
browse-page "https://example.com/article" --json

Decision Tree for Agents#

text

User asks a question
        │
        ▼
Is it purely creative/hypothetical?
   YES → Answer directly
   NO  ▼
Does it involve facts, events, versions, or data?
   YES ▼
Run: search-web "<query>" [--type news if recent event]
        │
        ▼
Are snippets sufficient to answer?
   YES → Synthesize answer + cite sources
   NO  ▼
Run: browse-page "<top_result_url>"
        │
        ▼
Synthesize answer from full page content + cite source

Why Search-First?#

Default LLM BehaviorSearch-First BehaviorAnswers from training dataAnswers from live webKnowledge cutoff appliesAlways up-to-dateMay produce outdated factsCites verifiable sourcesSingle knowledge sourceMulti-source cross-validation

Integration#

This skill works with:

• OpenClaw — native skill integration
• Claude Desktop / Cursor — via MCP server ( free-web-search-mcp)
• LangChain — via Python tool wrappers
• OpenAI Function Calling — via JSON schema definitions

MCP Server Setup (Claude Desktop / Cursor)#

Add to your claude_desktop_config.json:

json

{
  "mcpServers": {
    "free-web-search": {
      "command": "free-web-search-mcp",
      "args": []
    }
  }
}

Install via pip from the GitHub repository:

bash

pip install git+https://github.com/wd041216-bit/free-web-search-ultimate.git

Requirements#

• Python 3.8+
• beautifulsoup4, lxml, ddgs, mcp>=1.1.2

License#

MIT-0 — Free to use, modify, and redistribute. No attribution required.