Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Neural search via Exa MCP for web, code, and company research. Use when the user needs web search, code examples, company intel, people lookup, or AI-powered deep research with Exa's neural search engine.
.claude/skills/affaan-m-exa-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-21 | ✓→✓ | = Same ✓ | -62% | 0% |
通过 Exa MCP 服务器实现网页内容、代码、公司和人物的神经搜索。
必须配置 Exa MCP 服务器。添加到 ~/.claude.json:
json"exa-web-search": { "command": "npx", "args": ["-y", "exa-mcp-server"], "env": { "EXA_API_KEY": "YOUR_EXA_API_KEY_HERE" } }
在 exa.ai 获取 API 密钥。 此仓库当前的 Exa 设置记录了此处公开的工具接口:web_search_exa 和 get_code_context_exa。 如果你的 Exa 服务器公开了其他工具,请在文档或提示中依赖它们之前,先核实其确切名称。
用于当前信息、新闻或事实的通用网页搜索。
web_search_exa(query: "2026年最新人工智能发展", numResults: 5)参数:
| 参数 | 类型 | 默认值 | 说明 | |-------|------|---------|-------| | query | 字符串 | 必填 | 搜索查询 | | numResults | 数字 | 8 | 结果数量 | | type | 字符串 | auto | 搜索模式 | | livecrawl | 字符串 | fallback | 需要时优先使用实时爬取 | | category | 字符串 | 无 | 可选焦点,例如 company 或 research paper |
从 GitHub、Stack Overflow 和文档站点查找代码示例和文档。
get_code_context_exa(query: "Python asyncio patterns", tokensNum: 3000)参数:
| 参数 | 类型 | 默认值 | 说明 | |-------|------|---------|-------| | query | string | 必需 | 代码或 API 搜索查询 | | tokensNum | number | 5000 | 内容令牌数(1000-50000) |
web_search_exa(query: "Node.js 22 新功能", numResults: 3)get_code_context_exa(query: "Rust错误处理模式Result类型", tokensNum: 3000)web_search_exa(query: "Vercel 2026年融资估值", numResults: 3, category: "company")
web_search_exa(query: "site:linkedin.com/in Anthropic AI安全研究员", numResults: 5)web_search_exa(query: "WebAssembly 组件模型状态与采用情况", numResults: 5)
get_code_context_exa(query: "WebAssembly 组件模型示例", tokensNum: 4000)web_search_exa 获取最新信息、公司查询和广泛发现site:、引号内的短语和 intitle: 等搜索运算符来缩小结果范围tokensNum (1000-2000);对于全面的上下文,使用较高的值 (5000+)get_code_context_exadeep-research — 使用 firecrawl + exa 的完整研究工作流market-research — 带有决策框架的业务导向研究| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 38,690 | 6,732 | -83% | 1 | 1 | 0% | 3,532 | 1,333 | -62% | 0 | 0 | — |
case-06 | fail→fail | 16,344 | 5,124 | -69% | 1 | 1 | 0% | 2,151 | 1,193 | -45% | 0 | 0 | — |
case-02 | fail→fail | 11,697 | 5,677 | -51% | 1 | 1 | 0% | 2,087 | 1,299 | -38% | 0 | 0 | — |
case-03 | fail→fail | 18,807 | 5,149 | -73% | 1 | 1 | 0% | 2,858 | 1,170 | -59% | 0 | 0 | — |
case-04 | fail→fail | 4,877 | 6,777 | +39% | 1 | 1 | 0% | 1,064 | 2,289 | +115% | 0 | 0 | — |
case-05 | fail→fail | 14,937 | 7,084 | -53% | 1 | 1 | 0% | 3,051 | 1,458 | -52% | 0 | 0 | — |
case-21 | pass→pass | 34,942 | 16,479 | -53% | 1 | 1 | 0% | 6,169 | 2,342 | -62% | 0 | 0 | — |
case-07 | fail→fail | 14,217 | 5,852 | -59% | 1 | 1 | 0% | 2,334 | 1,218 | -48% | 0 | 0 | — |
case-08 | fail→fail | 4,591 | 6,839 | +49% | 1 | 1 | 0% | 766 | 1,338 | +75% | 0 | 0 | — |
case-09 | fail→fail | 17,653 | 7,055 | -60% | 1 | 1 | 0% | 3,111 | 1,320 | -58% | 0 | 0 | — |
case-10 | fail→fail | 12,451 | 5,151 | -59% | 1 | 1 | 0% | 2,453 | 1,217 | -50% | 0 | 0 | — |
case-11 | fail→fail | 12,660 | 7,124 | -44% | 1 | 1 | 0% | 1,802 | 1,250 | -31% | 0 | 0 | — |
case-12 | fail→fail | 14,096 | 7,596 | -46% | 1 | 1 | 0% | 2,258 | 1,368 | -39% | 0 | 0 | — |
case-13 | fail→fail | 14,506 | 6,113 | -58% | 1 | 1 | 0% | 2,436 | 1,238 | -49% | 0 | 0 | — |
case-14 | fail→pass | 10,128 | 3,078 | -70% | 1 | 1 | 0% | 1,873 | 1,331 | -29% | 0 | 0 | — |
case-15 | pass→pass | 3,267 | 1,752 | -46% | 1 | 1 | 0% | 555 | 1,233 | +122% | 0 | 0 | — |
case-16 | fail→pass | 6,595 | 1,558 | -76% | 1 | 1 | 0% | 1,048 | 1,159 | +11% | 0 | 0 | — |
case-17 | fail→pass | 9,686 | 2,649 | -73% | 1 | 1 | 0% | 1,741 | 1,291 | -26% | 0 | 0 | — |
case-18 | pass→pass | 7,107 | 1,951 | -73% | 1 | 1 | 0% | 1,166 | 1,244 | +7% | 0 | 0 | — |
case-19 | fail→pass | 9,459 | 1,656 | -82% | 1 | 1 | 0% | 800 | 1,105 | +38% | 0 | 0 | — |
case-20 | pass→pass | 18,927 | 19,190 | +1% | 1 | 1 | 0% | 3,223 | 4,073 | +26% | 0 | 0 | — |
case-22 | pass→pass | 7,048 | 6,670 | -5% | 1 | 1 | 0% | 782 | 2,063 | +164% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 10 counted toward the lift figure. The other 12 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +18 percentage points is the difference between those two pass rates over the 10 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/3/2026 | +48% |
Other measured skills in the registry, with their headline benchmark lift.