Ishan Goswami

Web Search by Exa

知识管理v2.0.0
作者:Ishan Goswami更新于 2026-05-11共 3 个版本 · 最新版本发布于 2026-03-11web-search-exa@clawhub_theishangoswami/web-search-exa
信息检索#Web Search#Documentation#Search The Web#News

据 AlphaMoat 统计,截至 2026-10,Web Search by Exa 累计下载 45,823 次、获 53 个星标,按累计下载量在本次参与排名的 10,914 个 AI 技能中排名第 546 位,在「知识管理」分类(745 个)中排名第 38 位。

列表默认按累计下载量排序,也可按星标排序;详情页排名按累计下载量计算。

累计下载
4.6万
星标
53
全站下载量排名
#546
全站前 6% · 「知识管理」前 6%
平台累计下载星标评论
ClawHub ↗3.4万44-
SkillHub ↗4.6万53—

各平台的计数口径不同,统计时间不一致。

这个技能是做什么的

Exa MCP 按语义搜索网页,并返回适合模型使用的内容,用户无需自行抓取结果。可设置结果条数与 auto、fast、deep 模式;agent 流程可用 highlights 代替全文以减少 token 用量。API key 可提高限额并解锁全部工具。搜索提示可能经由 Exa 处理。不要把 key 放进共享 URL、配置文件、日志或截图;人物/个人资料搜索仅用于合法的公开资料研究,并先核查隐私与平台条款。来源标出可疑安全结论,安装前须复核。

资料发布于 2026-10-27

可以用来做什么

  • 按语义描述找到页面并提取相关内容
  • 为获授权的研究检索公开公司、人物或代码资料

什么时候用

  • 关键词检索效果差、需要按意思找页面时
  • 想让检索结果直接进入模型上下文时

示例提示词

以下是根据技能文档生成的示例,不代表实际运行结果。

帮我找讲这个技术方案细节的页面,取前几条结果
用高亮模式检索这个主题,只要相关段落

月度数据趋势

累计下载4.6万
2.6万3.7万4.7万2026-03 · 累计下载 2.8万2026-05 · 累计下载 3.1万2026-06 · 累计下载 3.1万2026-07 · 累计下载 3.3万2026-08 · 累计下载 3.3万2026-09 · 累计下载 3.5万2026-10 · 累计下载 4.6万26-0326-0926-10
星标53
4349542026-03 · 星标 442026-05 · 星标 492026-06 · 星标 522026-07 · 星标 532026-08 · 星标 532026-09 · 星标 532026-10 · 星标 5326-0326-0926-10

2026-09 起统计方法调整:同名技能按作者分别统计,并覆盖全部技能。图中虚线为调整前的月份,与之后的数据不直接比较,也不计算跨越调整的增长。

月份累计下载环比星标
2026-104.6万—53
2026-093.5万—53
2026-083.3万+1.3%53
2026-073.3万+3.7%53
2026-063.1万+1.7%52
2026-053.1万—49
2026-032.8万—44

SKILL.md 原文

以下是作者发布的原始文件,按纯文本显示,本站未做修改或审核;其中的链接不可点击,命令请核实后再运行。抓取于 2026-10-23。

11 KB · ClawHub v2.0.0
---
name: web-search-exa
description: "Neural web search, content extraction, company and people research, code search, and deep research via the Exa MCP server. Use when you need to: (1) search the web with semantic understanding — not just keywords, (2) find research papers, news, tweets, companies, or people, (3) extract clean content from URLs, (4) find semantically similar pages to a known URL, (5) get code examples and documentation, (6) run deep multi-step research with a report, (7) get a quick synthesized answer with citations. NOT for: local file operations, non-web tasks, or anything that doesn't involve web search or content retrieval."
---

# Exa — Neural Web Search & Research

Exa is a neural search engine. Unlike keyword-based search, it understands meaning — you describe the page you're looking for and it finds it. Returns clean, LLM-ready content with no scraping needed.

**MCP server:** `https://mcp.exa.ai/mcp`
展开全部(共 358 行)
**Free tier:** generous rate limits, no key needed for basic tools
**API key:** [dashboard.exa.ai/api-keys](https://dashboard.exa.ai/api-keys) — unlocks higher limits + all tools
**Docs:** [exa.ai/docs](https://exa.ai/docs)
**GitHub:** [github.com/exa-labs/exa-mcp-server](https://github.com/exa-labs/exa-mcp-server)

## Setup

Add the MCP server to your agent config:

```bash
# OpenClaw
openclaw mcp add exa --url "https://mcp.exa.ai/mcp"
```

Or in any MCP config JSON:
```json
{
  "mcpServers": {
    "exa": {
      "url": "https://mcp.exa.ai/mcp"
    }
  }
}
```

To unlock all tools and remove rate limits, append your API key:
```
https://mcp.exa.ai/mcp?exaApiKey=YOUR_EXA_KEY
```

To enable specific optional tools:
```
https://mcp.exa.ai/mcp?exaApiKey=YOUR_KEY&tools=web_search_exa,web_search_advanced_exa,people_search_exa,crawling_exa,company_research_exa,get_code_context_exa,deep_researcher_start,deep_researcher_check,deep_search_exa
```

---

## Tool Reference

### Default tools (available without API key)

| Tool | What it does |
|------|-------------|
| `web_search_exa` | General-purpose web search — clean content, fast |
| `get_code_context_exa` | Code examples + docs from GitHub, Stack Overflow, official docs |
| `company_research_exa` | Company overview, news, funding, competitors |

### Optional tools (enable via `tools` param, need API key for some)

| Tool | What it does |
|------|-------------|
| `web_search_advanced_exa` | Full-control search: domain filters, date ranges, categories, content modes |
| `crawling_exa` | Extract full page content from a known URL — handles JS, PDFs, complex layouts |
| `people_search_exa` | Find LinkedIn profiles, professional backgrounds, experts |
| `deep_researcher_start` | Kick off an async multi-step research agent → detailed report |
| `deep_researcher_check` | Poll status / retrieve results from deep research |
| `deep_search_exa` | Single-call deep search with synthesized answer + citations (needs API key) |

---

## web_search_exa

Fast general search. Describe what you're looking for in natural language.

**Parameters:**
- `query` (string, required) — describe the page you want to find
- `numResults` (int) — number of results, default 10
- `type` — `auto` (best quality), `fast` (lower latency), `deep` (multi-step reasoning)
- `livecrawl` — `fallback` (default) or `preferred` (always fetch fresh)
- `contextMaxCharacters` (int) — cap the returned content size

```
web_search_exa {
  "query": "blog posts about using vector databases for recommendation systems",
  "numResults": 8
}
```

```
web_search_exa {
  "query": "latest OpenAI announcements March 2026",
  "numResults": 5,
  "type": "fast"
}
```

---

## web_search_advanced_exa

The power-user tool. Everything `web_search_exa` does, plus domain filters, date filters, category targeting, and content extraction modes.

**Extra parameters beyond basic search:**

| Parameter | Type | What it does |
|-----------|------|-------------|
| `includeDomains` | string[] | Only return results from these domains (max 1200) |
| `excludeDomains` | string[] | Block results from these domains |
| `category` | string | Target content type — see table below |
| `startPublishedDate` | string | ISO date, results published after this |
| `endPublishedDate` | string | ISO date, results published before this |
| `maxAgeHours` | int | Content freshness — `0` = always livecrawl, `-1` = cache only, `24` = cache if <24h |
| `contents.highlights` | object | Extractive snippets relevant to query. Set `maxCharacters` to control size |
| `contents.text` | object | Full page as clean markdown. Set `maxCharacters` to cap |
| `contents.summary` | object | LLM-generated summary. Supports `query` and JSON `schema` for structured extraction |

**Categories:**

| Category | Best for |
|----------|---------|
| `company` | Company pages, LinkedIn company profiles |
| `people` | LinkedIn profiles, professional bios, personal sites |
| `research paper` | arXiv, academic papers, peer-reviewed research |
| `news` | Current events, journalism |
| `tweet` | Posts from X/Twitter |
| `personal site` | Blogs, personal pages |
| `financial report` | SEC filings, earnings reports |

### Examples

**Research papers:**
```
web_search_advanced_exa {
  "query": "transformer architecture improvements for long-context windows",
  "category": "research paper",
  "numResults": 15,
  "contents": { "highlights": { "maxCharacters": 3000 } }
}
```

**Company list building with structured extraction:**
```
web_search_advanced_exa {
  "query": "Series A B2B SaaS companies in climate tech founded after 2022",
  "category": "company",
  "numResults": 25,
  "contents": {
    "summary": {
      "query": "company name, what they do, funding stage, location",
      "schema": {
        "type": "object",
        "properties": {
          "name": { "type": "string" },
          "description": { "type": "string" },
          "funding": { "type": "string" },
          "location": { "type": "string" }
        }
      }
    }
  }
}
```

**People search — find candidates with specific profiles:**
```
web_search_advanced_exa {
  "query": "machine learning engineers at fintech startups in NYC with experience in fraud detection",
  "category": "people",
  "numResults": 20,
  "contents": { "highlights": { "maxCharacters": 2000 } }
}
```

**Finding pages similar to a known URL:**
Use the URL itself as the query — Exa will find semantically similar pages:
```
web_search_advanced_exa {
  "query": "https://linkedin.com/in/some-candidate-profile",
  "numResults": 15,
  "contents": { "highlights": { "maxCharacters": 2000 } }
}
```

**Recent news with freshness control:**
```
web_search_advanced_exa {
  "query": "AI regulation policy updates",
  "category": "news",
  "maxAgeHours": 72,
  "numResults": 10,
  "contents": { "highlights": { "maxCharacters": 4000 } }
}
```

**Scoped domain search:**
```
web_search_advanced_exa {
  "query": "authentication best practices",
  "includeDomains": ["owasp.org", "auth0.com", "docs.github.com"],
  "numResults": 10,
  "contents": { "text": { "maxCharacters": 5000 } }
}
```

---

## company_research_exa

One-call company research. Returns business overview, recent news, funding, and competitive landscape.

```
company_research_exa { "query": "Stripe payments company overview and recent news" }
```

```
company_research_exa { "query": "what does Anduril Industries do and who are their competitors" }
```

---

## people_search_exa

Find professionals by role, company, location, expertise. Returns LinkedIn profiles and bios.

```
people_search_exa { "query": "VP of Engineering at healthcare startups in San Francisco" }
```

```
people_search_exa { "query": "AI researchers specializing in multimodal models" }
```

---

## get_code_context_exa

Search GitHub repos, Stack Overflow, and documentation for code examples and API usage patterns.

```
get_code_context_exa { "query": "how to implement rate limiting in Express.js with Redis" }
```

```
get_code_context_exa { "query": "Python asyncio connection pooling example with aiohttp" }
```

---

## crawling_exa

Extract clean content from a specific URL. Handles JavaScript-rendered pages, PDFs, and complex layouts. Returns markdown.

```
crawling_exa { "url": "https://arxiv.org/abs/2301.07041" }
```

Good for when you already have the URL and want to read the page.

---

## deep_researcher_start + deep_researcher_check

Long-running async research. Exa's research agent searches, reads, and compiles a detailed report.

**Start a research task:**
```
deep_researcher_start {
  "query": "competitive landscape of AI code generation tools in 2026 — key players, pricing, technical approaches, market share"
}
```

**Check status (use the researchId from the start response):**
```
deep_researcher_check { "researchId": "abc123..." }
```

Poll `deep_researcher_check` until status is `completed`. The final response includes the full report.

---

## deep_search_exa

Single-call deep search: expands your query across multiple angles, searches, reads results, and returns a synthesized answer with grounded citations. Requires API key.

```
deep_search_exa { "query": "what are the leading approaches to multimodal RAG in production systems" }
```

Supports structured output via `outputSchema`:
```
deep_search_exa {
  "query": "top 10 aerospace companies by revenue",
  "type": "deep",
  "outputSchema": {
    "type": "object",
    "required": ["companies"],
    "properties": {
      "companies": {
        "type": "array",
        "items": {
          "type": "object",
          "properties": {
            "name": { "type": "string" },
            "revenue": { "type": "string" },
            "hq": { "type": "string" }
          }
        }
      }
    }
  }
}
```

---

## Query Craft

Exa is neural — it matches on meaning, not keywords. Write queries like you'd describe the ideal page to a colleague.

**Do:** "blog post about using embeddings for product recommendations at scale"
**Don't:** "embeddings product recommendations"

**Do:** "Stripe payments company San Francisco fintech"
**Don't:** "Stripe" (too ambiguous)

- Use `category` when you know the content type — it makes a big difference.
- For broader coverage, run 2-3 query variations in parallel and deduplicate results.
- For agentic workflows, use `highlights` instead of full `text` — it's 10x more token-efficient while keeping the relevant parts.

## Token Efficiency

| Content mode | When to use |
|-------------|------------|
| `highlights` | Agent workflows, factual lookups, multi-step pipelines — most token-efficient |
| `text` | Deep analysis, when you need full page context |
| `summary` | Quick overviews, structured extraction with JSON schema |

Set `maxCharacters` on any content mode to control output size.

## When to Reach for Which Tool

| I need to... | Use |
|-------------|-----|
| Quick web lookup | `web_search_exa` |
| Research papers, academic search | `web_search_advanced_exa` + `category: "research paper"` |
| Company intel, competitive analysis | `company_research_exa` or advanced + `category: "company"` |
| Find people, candidates, experts | `people_search_exa` or advanced + `category: "people"` |
| Code examples, API docs | `get_code_context_exa` |
| Read a specific URL | `crawling_exa` |
| Find pages similar to a URL | `web_search_advanced_exa` with URL as query |
| Recent news / tweets | Advanced + `category: "news"` or `"tweet"` + `maxAgeHours` |
| Detailed research report | `deep_researcher_start` → `deep_researcher_check` |
| Quick answer with citations | `deep_search_exa` |

---

**Docs:** [exa.ai/docs](https://exa.ai/docs) — **Dashboard:** [dashboard.exa.ai](https://dashboard.exa.ai) — **Support:** [email protected]

「知识管理」分类下的热门技能

搜索引擎知识管理

通过公开搜索URL在16个搜索引擎(7个国内直连+9个国际)中做跨引擎检索、结果去重与多源交叉验证,无需API Key,内置零依赖Python脚本。Use when 用户要求“搜一下/查资料/多找几个来源/核实消息/找公众号文章/排查报错/找论文或开源项目”时。输入自然语言查询词,输出带来源链接与可信度标注的结构化结果。不适用于私有数据检索、批量抓取、需要登录后可见的内容。

↓ 1075.9万★ 147u_8ee34d77
ima-skills知识管理

通过 IMA OpenAPI 管理笔记与知识库内容。

↓ 108.4万★ 566u_1742680f
Summarize知识管理

把长文本、文档与网页内容压缩成要点和关键词的摘要工具。

↓ 87.6万★ 1,048paudyyin
全网搜索知识管理

通过“查一下、搜一下、搜索、找一下”等自然语言检索公开网页,整理来源并交付横版 PDF、Markdown 与标题链接表。支持用户主动开启、查看和取消技能收藏,保留原报告取回。

↓ 50.9万★ 2u_32d99f8a
web-tools-guide知识管理

规定如何为网页请求选择合适工具的必读指引。

↓ 40.0万★ 220user_ec205dbb
Aihot Skill Lite知识管理

AI HOT(aihot.virxact.com)中文 AI 资讯查询 Skill,帮助用户获取今日 AI 圈动态、AI 日报、AI 热点、最新 AI 资讯以及 OpenAI、Anthropic、Google 等公司的最新发布信息。

↓ 32.1万★ 92Khazix
全球12亿文献知识库(8千万中文期刊可下载)知识管理

全球12亿文献知识库(8千万中文期刊可下载)——通过 SmartLib 开放平台 API 提供中外文学术文献检索与下载能力,覆盖 8000 万篇授权中文期刊全文 + 12.28 亿条全球文献元数据(期刊 7.19 亿 / 专利 2.15 亿 / 会议 7155 万 / 学位论文 2473 万 / 标准 268 万)。 能力:中英文关键词检索、文献详情、中文期刊 PDF 全文下载、外文 OA 文献十级渠道免费下载(不消耗配额)、智能关键词扩展、核心期刊优先排序、相关性重排、引文追溯、分类号检索。 配额:首次使用自动注册,免费 100 次检索 + 10 次下载 / 月;耗尽自动弹出套餐(体验卡 / 个人版月 / 专业版月 / 单篇下载 / 下载包),可说「升级 / 充值」唤起;企业 / 机构定制联系 [email protected]。 触发:用户表达"查论文""找文献""检索学术""搜期刊""查专利""找标准""下论文""写文献综述""找参考文献""查 SCI/EI"等意图时启用;也适用"帮我找关于 XX 的论文""写文献综述""找几篇引用支撑论点"。英文:"find papers" "search literature" "write literature review" "find supporting citations"。 调用前必须先用 /consume 获取 consume_token,再凭 token 调 /search(每次计费接口调用都需一次 /consume)。

↓ 30.3万★ 189user_164f4c1f
Markdown Converter知识管理

用 uvx 直接调用 markitdown,把各类文件转成 Markdown。

↓ 20.4万★ 256Peter Steinberger

数据月份:2026-10 · 下载/星标数聚合自公开技能注册库(ClawHub、SkillHub),按月更新。口径说明见数据方法。