Web Search Pro
Multi-engine web search with full parameter control for AI agents.
A precision supplement to OpenClaw's built-in web_search (Brave/Perplexity), providing
domain filtering, deep search, news mode, date ranges, and content extraction
that the built-in search does not support.
多引擎精细化网络搜索，为 AI Agent 设计。
作为 OpenClaw 内置 web_search（Brave/Perplexity）的精细化补充，提供域名过滤、
深度搜索、新闻模式、日期范围、内容提取等内置搜索不支持的能力。
配置一个或多个 API Key，自动选择最优引擎。
Engines / 引擎
| Engine | Strengths | Free Tier | API Key Env |
|---|---|---|---|
| Tavily | AI-optimized, best answer quality, full filters, extract | 1000/month | TAVILY_API_KEY |
| Exa | Semantic/neural search, deep research | $10 credit | EXA_API_KEY |
| Serper | Real Google SERP, best news coverage | 100/month | SERPER_API_KEY |
| SerpAPI | Multi-engine (Google/Bing/Baidu/Yandex/DuckDuckGo) | 250/month | SERPAPI_API_KEY |
Auto-Select Priority / 自动选择优先级
When --engine is not specified, the skill picks the best available engine:
| Query Type | Priority | Reason |
|---|---|---|
| Default | Tavily > Exa > Serper > SerpAPI | Tavily has best AI answer + full filters |
--deep |
Tavily > Exa | Both have dedicated deep search modes |
--news |
Serper > Tavily | Google News has widest coverage |
--include-domains |
Tavily > Exa > Serper > SerpAPI | Tavily/Exa have native domain filters |
--search-engine baidu |
SerpAPI | Only SerpAPI supports Baidu/Yandex |
| Chinese queries | SerpAPI (Baidu) > Serper | Baidu has better Chinese results |
Search / 搜索
# Basic search (auto-select engine)
node {baseDir}/scripts/search.mjs "query"
# Force specific engine
node {baseDir}/scripts/search.mjs "query" --engine tavily
# Domain filtering (only search specific sites)
node {baseDir}/scripts/search.mjs "query" --include-domains "github.com,stackoverflow.com"
# Exclude domains
node {baseDir}/scripts/search.mjs "query" --exclude-domains "pinterest.com,quora.com"
# Date range (absolute)
node {baseDir}/scripts/search.mjs "query" --from 2026-01-01 --to 2026-02-09
# Time range (relative)
node {baseDir}/scripts/search.mjs "query" --time-range week
# Deep/advanced search
node {baseDir}/scripts/search.mjs "query" --deep
# News search
node {baseDir}/scripts/search.mjs "query" --news --days 7
# Multi-engine: Baidu search
node {baseDir}/scripts/search.mjs "query" --engine serpapi --search-engine baidu
# More results
node {baseDir}/scripts/search.mjs "query" -n 10
# JSON output (for programmatic use)
node {baseDir}/scripts/search.mjs "query" --json
Extract / 内容提取
Extract readable content from URLs (Tavily Extract or Exa livecrawl):
node {baseDir}/scripts/extract.mjs "https://example.com/article"
node {baseDir}/scripts/extract.mjs "url1" "url2" "url3"
node {baseDir}/scripts/extract.mjs "url" --engine exa
node {baseDir}/scripts/extract.mjs "url" --json
All Options / 全部参数
| Option | Description | Engines |
|---|---|---|
--engine <name> |
Force engine: tavily\ | exa\ |
-n <count> |
Number of results (default: 5) | all |
--deep |
Deep/advanced search mode | tavily, exa |
--news |
News search mode | tavily, serper, serpapi |
--days <n> |
Limit news to last N days | tavily |
--include-domains <d,...> |
Only search these domains | all (native: tavily, exa) |
--exclude-domains <d,...> |
Exclude these domains | all (native: tavily, exa) |
--time-range <range> |
day\ | week\ |
--from <YYYY-MM-DD> |
Results after this date | all |
--to <YYYY-MM-DD> |
Results before this date | all |
--search-engine <name> |
SerpAPI sub-engine: google\ | bing\ |
--country <code> |
Country code (us, cn, de...) | serper, serpapi |
--lang <code> |
Language code (en, zh, de...) | serper, serpapi |
--json |
Raw JSON output | all |
Setup / 配置
Add API keys to your environment (e.g., ~/.openclaw/.env):
# Configure at least one (recommended: Tavily)
TAVILY_API_KEY=tvly-xxxxx # https://tavily.com (1000 free/month)
EXA_API_KEY=exa-xxxxx # https://exa.ai ($10 free credit)
SERPER_API_KEY=xxxxx # https://serper.dev (100 free/month)
SERPAPI_API_KEY=xxxxx # https://serpapi.com (250 free/month)
Notes / 说明
- At least one API key must be configured
- Domain filtering via
--include-domains/--exclude-domainsworks natively on Tavily and Exa; on Serper/SerpAPI it's implemented viasite:query operators --deepmode uses more API credits (Tavily: 2x, Exa: varies)- Extract only works with Tavily and Exa
- All output is Markdown-formatted for AI agent consumption; use
--jsonfor programmatic access