leo-tavily-mcp

leo-tavily-mcp

Enables Leo's web_search tool to perform live web searches using the Tavily API, returning structured results as JSON. Configurable via a single API key, and pinned to a commit SHA for secure distribution.

Category
访问服务器

README

leo-tavily-mcp

Tavily web search as a Leo provider package, over MCP.

Leo's web_search tool doesn't search anything itself — it resolves a WebSearchProvider and asks that. This server is one, reachable as a package the hub installs at runtime rather than one it has to be rebuilt for.

What it answers

The web_search role in leo_mcp::providers:

search({ query, max_results }) -> { results: [{ title, url, snippet, score? }] }

Results come back as JSON in a text content block, which is the convention the hub's bridge reads.

Configuration

One setting, declared by the package descriptor and handed to this process by Leo as an environment variable under its settings key, verbatim:

env where it comes from
tavily_api_key Settings → Packages → Web Search

Leo only hands over keys the package declared in entitlements.settings_read and the owner consented to — never the whole settings table. The name is lower-case because it is the settings key; renaming it on either side means the credential silently never arrives.

With no key configured the tool returns isError with a message naming the setting. That is deliberate: an empty result set would reach Leo as "the web has nothing on that".

Running it

npm install
node index.js          # speaks MCP over stdio
npm test               # checks the response mapping, no network needed

Why plain JavaScript

This ships as a git tarball pinned to a commit SHA rather than an npm package, and npm does not reliably run build steps for a tarball URL. A TypeScript source tree would install and then fail to start on somebody else's machine, at the far end, where it is hardest to diagnose.

Registry entry

{
  "name": "tavily",
  "label": "Web Search",
  "mcp": {
    "transport": "stdio",
    "command": "npx",
    "args": ["-y", "https://github.com/TheBananaStand/leo-tavily-mcp/archive/<40-char-sha>.tar.gz"]
  },
  "descriptor": {
    "provides": ["web_search"],
    "fields": [
      { "key": "tavily_api_key", "label": "API Key", "field_type": "password", "required": true }
    ],
    "entitlements": {
      "settings_read": ["tavily_api_key"],
      "network": ["api.tavily.com"],
      "sends": [
        { "to": "api.tavily.com", "classes": ["search_queries"],
          "because": "The search text and nothing else." }
      ]
    }
  }
}

The URL must name a full 40-character commit SHA. A tag looks equally specific and is not — refs/tags/v1.0.0 is a pointer its owner can move after the entry has been reviewed, so the hub rejects it as unpinned.

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选