A web search tool based on the Google Serper API, providing real-time, accurate search results.
When to Activate
Automatically activated when the user mentions the following:
Search-type keywords
- "搜索", "搜一下", "搜搜", "查找", "找一下"
- "研究", "调研", "了解"
- "查询", "检索"
- "看看", "查查"
Specific scenarios
- Need to get the latest news and information
- Need to verify facts or data
- Need to research a certain technology or concept
- Need to find documentation and tutorials
- Need to compare different products or solutions
Example questions
- "搜一下最新的 AI 发展趋势"
- "帮我搜索最新的 AI 发展趋势"
- "查找一下 Python 3.13 的新特性"
- "研究一下自动驾驶技术的现状"
- "查查最新的网络安全新闻"
- "找一些关于微服务的教程"
Tools
serper_search
Purpose: Perform a web search and return a list of results
Parameters:
query(required, string): search keywordsnum(optional, number): number of results to return, default 5, maximum 20gl(optional, string): country code, default cn- Recommended values: cn (China), us (United States), uk (United Kingdom), jp (Japan)
hl(optional, string): language code, default zh-CN- Recommended values: zh-CN (Simplified Chinese), en (English), ja (Japanese)
Best Practices
1. Search techniques
Use specific keywords and avoid being too broad:
Example:
- ✅ "Kimi AI 模型 参数 对比 2025"
- ✅ "Python 3.13 新特性 官方文档"
- ❌ "Python" (too broad, too many results)
2. Add time constraints
Specify the time range to get the latest information:
Example:
- ✅ "LangChain 最新文档 2025"
- ✅ "Python 3.13 发布时间"
- ✅ "AI 人工智能 新闻 2025年2月"
3. Use exact search
Use quotation marks to search for exact phrases:
Example:
- ✅ ""machine learning" 最佳实践"
- ✅ ""RAG 架构" 实现"
4. Add technical terms
Improve search precision:
Example:
- ✅ "Spring Cloud 微服务 实现"
- ✅ "React Hooks useEffect 使用"
5. Choosing the number of results
Adjust according to your needs:
- Quick lookup:
num=3(quickly scan the core information) - In-depth research:
num=10(comprehensively understand the topic) - Comprehensive research:
num=20(maximum depth, broad coverage)
6. Multi-round search strategy
For complex topics, use multiple rounds of search to dig deeper:
Scenario: researching microservice architecture
// 第一轮:基础概念
serper_search({ query: "微服务架构 概念 优势", num: 5 })
// 第二轮:技术实现
serper_search({ query: "微服务 Spring Cloud Kubernetes", num: 8 })
// 第三轮:最佳实践
serper_search({ query: "微服务 最佳实践 常见问题", num: 10 })
7. Principles for handling results
- Prioritize authoritative sources: Official documentation, well-known tech blogs, academic papers
- Cross-verification: Compare information from multiple sources
- Pay attention to timeliness: Prioritize more recent content (within the last 1-2 years)
- Quick filtering: Judge relevance by snippet
Example Scenarios
Scenario 1: Technical research
User question: "Research the core concepts of LangChain"
AI behavior:
// 搜索核心概念
serper_search({
query: "LangChain 核心概念 官方文档 2025",
num: 5
})
// 如果需要更多细节,继续搜索
serper_search({
query: "LangChain Agents Chains 使用示例",
num: 8
})
Example response:
根据搜索结果,LangChain 的核心概念包括:
1. **Chains(链)**:将多个组件串联成工作流
2. **Agents(智能体)**:自主决策的 AI 代理
3. **Tools(工具)**:LLM 可用的外部能力
4. **Memory(记忆)**:对话上下文管理
📖 详细文档:https://python.langchain.com/
Scenario 2: News retrieval
User question: "What AI news is there today?"
AI behavior:
serper_search({
query: "AI 人工智能 新闻 2025年2月6日",
num: 10
})
Example response:
今日 AI 要闻:
1. **OpenAI 发布新模型**:能力提升 30%
2. **Google DeepMind**:突破蛋白质折叠新算法
3. **中国 AI 监管**:新规发布,促进行业健康发展
🔍 来源:各大科技媒体报道
Scenario 3: Product comparison
User question: "Which is better, Kimi or ChatGPT?"
AI behavior:
// 搜索对比评测
serper_search({
query: "Kimi ChatGPT 对比评测 2025",
num: 5
})
Example response:
根据评测对比:
| 特性 | Kimi | ChatGPT |
|------|------|---------|
| 中文理解 | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ |
| 工具调用 | ⭐⭐⭐ | ⭐⭐⭐⭐⭐ |
| 响应速度 | 快 | 中等 |
💡 建议:中文任务用 Kimi,复杂工具调用用 ChatGPT
Scenario 4: Fact verification
User question: "I heard Python 4 is about to be released. Is that true?"
AI behavior:
serper_search({
query: "Python 4 发布计划 官方声明",
num: 5
})
Example response:
经核实:
❌ Python 4 目前没有发布计划
✅ Python 3.13 是最新版本(2024年10月发布)
✅ Python 3.14 计划于 2025 年发布
📊 信息来源:Python 官方 GitHub、PSF 公告
Scenario 5: Quick lookup
User question: "Help me find a React Hooks tutorial"
AI behavior:
serper_search({
query: "React Hooks 官方教程 中文",
num: 3
})
Example response:
推荐教程:
1. 📚 **React 官方文档**(推荐)
https://react.dev/learn
2. 🎓 **阮一峰 React Hooks 教程**
https://www.ruanyifeng.com/blog/2019/09/react-hooks.html
3. 📺 **B站 React Hooks 视频教程**
搜索"React Hooks 入门"
Limitations
- Search result source: Google, may be affected by region and personalization
- Access restrictions: Some websites may be blocked or require login
- Number of results: Up to 20
- Rate limit: The API has call frequency limits; avoid a large number of requests in a short time
- Search quality: Depends on the accuracy and timing of keywords
Configuration
Environment variable configuration (recommended)
Edit ~/.openclaw/.env:
SERPER_API_KEY=your-api-key-here
Plugin configuration
In openclaw.plugin.json 's configSchema , configure apiKey。
Get API Key
Visit https://serper.dev/ to register and get an API Key.
Free quota: 2,500 calls per month.
Related Tools
- web_search: Brave Search API (alternative option)
- web_fetch: Get detailed content of a single web page
Version History
-
v1.0 (2026-02-06): Initial version, basic search functionality
- Support Google Serper API
- Provide Chinese and English multilingual search
- Integrate with the OpenClaw Skill system
-
v1.1 (2026-02-11): Optimize description, add trigger keywords
- Update description to include trigger keywords
- Optimize the skill activation mechanism
Future Plans
- 📚 serper-scholar: Academic search support
- 🔍 Advanced filtering: Time range, domain filtering
- 📊 Result caching: Improve performance for repeated queries
- 🌍 More regions: Support more countries and languages
Tips: When using search, try to express your needs in natural language, and the AI will automatically help you construct an appropriate search query.