Skip to main content
Glama
exa-mcp-server-for-operit

exa-mcp-server-for-operit

by x15907982411

Exa MCP Server(Operit 适配版)

version license

本仓库是 exa-labs/exa-mcp-server v3.4.0 的 Operit 平台适配版。

原版 Exa MCP Server 用于将 Exa 的搜索能力(web search / code search / company research)接入 AI 助手;本适配版在保持全部功能不变的前提下,修复了 Exa MCP 与 Operit 平台(旧版 Kotlin MCP SDK / Zod schema 校验)握手失败的问题,使 web_search_exa / web_fetch_exa 等工具可在 Operit 中稳定加载与调用。


为什么需要适配(背景)

Operit 使用的 MCP 客户端 SDK 版本较旧,其 Zod 校验要求 serverInfo 响应中不允许出现 icons 字段:

现象

原因

MCP 插件一直"卡在加载"

客户端发起 initialize 后,服务端返回的 serverInfo 携带 icons,旧 SDK 校验报 invalid_type,握手失败

工具列表拉不到 / 工具不可用

握手未完成,tools/list 无法正常返回

本仓库将 src/stdio.ts 中 McpServer 构造参数的 icons 字段移除,并使源码与构建产物(dist/stdio.cjs)保持一致,构建可复现。

注:dist/stdio.cjs 中仍保留 SDK 库内部的 IconSchema/IconsSchema 定义(用于校验客户端传入字段的 schema 声明),这与 serverInfo 响应中的 icons 无关,不影响兼容性。

Related MCP server: Exa MCP Server

与原仓库的差异(全部)

文件

修改

说明

src/stdio.ts

移除 McpServer 的 icons 参数

兼容 Operit 旧 SDK 的 Zod 校验(修复根因)

src/stdio.ts / package.json

版本号 3.4.0 → 3.4.0-operit.1

发布版本可识别,serverInfo 回显该版本

除此之外,工具集、API、skills、测试、Dockerfile、vercel 配置等均与上游 v3.4.0 完全一致。

快速开始

1. 构建(可复现)

npm ci                    # 安装依赖(国内可换 npmmirror 加速:npm config set registry https://registry.npmmirror.com)
npm run build:stdio       # esbuild 打包 → dist/stdio.cjs(自包含,约 2.2MB)

⚠️ npm 11+ 默认阻止 postinstall 脚本,若报 esbuild: command not found,执行:

npm approve-scripts esbuild   # 或 npm approve-builds
node node_modules/esbuild/install.js

2. 验证握手(initialize)

node dist/stdio.cjs <<'EOF'
{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}
EOF

预期 serverInfo 响应(无 icons 字段):

{
  "name": "exa-search-server",
  "title": "Exa",
  "version": "3.4.0-operit.1",
  "websiteUrl": "https://exa.ai"
}

3. 部署到 Operit

  1. 将本仓库(或仅 dist/ 目录)放入 Operit 的 MCP 插件目录:

    • Android 源目录:/sdcard/Download/Operit/mcp_plugins/<pluginId>/

    • Linux 运行目录:~/mcp_plugins/<shortName>/(含 dist/stdio.cjs)

  2. 在 mcp_config.json 注册(示例):

{
  "mcpServers": {
    "exa": {
      "command": "node",
      "args": ["dist/stdio.cjs"],
      "env": {
        "EXA_API_KEY": "<your-exa-api-key>",
        "EXA_TOOL_ENABLES": "web_search_exa,web_fetch_exa"
      }
    }
  }
}
  1. 重启 MCP 服务(Operit 内触发 restart_mcp_with_logs),验证 ping_mcp 可列出工具。

可用工具

  • web_search_exa — 默认启用,Exa 语义搜索

  • web_fetch_exa — 默认启用,任意 URL 转 Markdown

  • web_search_advanced_exa 等高级工具 — 需在 EXA_TOOL_ENABLES 中追加工具名后启用

详见上游 Full Documentation。

详细适配说明

构建细节、踩坑记录与适配前后对比见 README-OPERIT.md。

上游信息

License

MIT — 保留上游 exa-labs/exa-mcp-server 版权声明,详见 LICENSE。

Available Tools

2 tools
web_fetch_exaA
Read-onlyIdempotent

Read a webpage's full content as clean markdown. Use after web_search_exa when highlights are insufficient or to read any URL.

Best for: Extracting full content from known URLs. Batch multiple URLs in one call. Returns: Clean text content and metadata from the page(s).

Inputs: urls, maxCharacters.

web_search_exaA
Read-onlyIdempotent

Search the web for any topic and get clean, ready-to-use content.

  Best for: Finding current information, news, facts, people, companies, or answering questions about any topic.
  Returns: Clean text content from top search results.

  Query tips:
  describe the ideal page, not keywords. "blog post comparing React and Vue performance" not "React vs Vue".
  Use category:people / category:company to search through Linkedin profiles / companies respectively.
  If highlights are insufficient, follow up with web_fetch_exa on the best URLs.

Inputs: query, numResults.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.5/5.0

    Scored across 2 tools

    Disambiguation5/5

    web_search_exa and web_fetch_exa have clearly distinct purposes: one searches the web and returns highlights, the other fetches full page content. There is no overlap in functionality, and their complementary nature is explicitly described.

    Naming Consistency5/5

    Both tools follow the same verb_noun pattern with a consistent shared prefix: web_search_exa and web_fetch_exa. The naming is predictable and reflects the action each tool performs.

    Tool Count4/5

    Two tools is minimal but appropriate for a server focused on web search and retrieval. The pair forms a complete workflow (search then fetch), and adding more tools for a simple utility like this could be overkill. The count feels slightly thin but not insufficient for the stated purpose.

    Completeness4/5

    The server covers the core lifecycle of web content retrieval: searching for relevant pages and fetching full content from known URLs. Minor gaps exist, such as no findSimilar or advanced filtering, but the essential use case is fully supported.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    • Web search API and MCP server for developers building AI agents.

      https://searchpipe.tech/mcp/
    • Your agent needs the open web — searched by more than one engine, and read as clean markdown rather than raw HTML. **What you can ask for** • "Search this question with two providers and tell me where they disagree." • "Scrape these 40 URLs into markdown, in one batch." • "Crawl this documentation site and give me every page." • "Do deep research on this topic and cite the sources." • "Find the academic papers behind this claim." **How to use it** Point any MCP client at https://mcp.aisa.one/search/mcp and sign in with OAuth — there is no key to create or paste. 30 tools across several independent providers: Tavily and Exa search, answers, contents and agent runs; Firecrawl scrape, batch scrape, crawl, map and search; Perplexity Sonar, Sonar Pro, reasoning and deep research; Oxylabs AI search and LLM jobs; OpenAI and Anthropic web search; and scholarly search. **Why this rather than the source** Several independent indexes behind one account, because one engine's blind spot is not visible from inside it. **It is also a door to the rest** The same login reaches 26 sources and 580+ operations. Find the page here, then ask the same agent who links to it or how much traffic it gets — without adding a second server. **What it costs** Finding and inspecting an operation is free. Running one is billed per call at API prices, with no seat and no monthly minimum, and every call takes max_price_usd so an agent cannot overspend by accident. **Where else it reaches** https://mcp.aisa.one/seo-serp/mcp for the Google results page itself, https://mcp.aisa.one/seo-serp-other-engines/mcp for Bing, Baidu and Naver.

    • Free web search for AI agents. No API key required. Hosted MCP in active development.

      https://api.ssearch.cc/mcp
    • The Remote MCP server acts as a standardized bridge between LLM applications (like Claude, ChatGPT, and Cursor) and external services, enabling AI agents to access external tools and resources. Its primary capability is providing a centralized search tool to discover other MCP servers and their respective tools. Unlike local implementations, it runs remotely with OAuth authentication and permission controls for security.

      https://mcp.remote-mcp.com

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      An MCP server that integrates real-time web search capabilities into AI assistants using the Exa API, providing both basic and advanced search functionality with formatted markdown results.
      144
      MIT
    • A
      license
      A
      quality
      D
      maintenance
      A Model Context Protocol server that enables AI assistants like Claude to perform real-time web searches using the Exa AI Search API in a safe and controlled manner.
      2
      46,492 npm
      MIT
    • A
      license
      B
      quality
      D
      maintenance
      Exposes the Exa web search API as an MCP server, providing tools for web searching, content retrieval, and similar page discovery. It also supports advanced features like AI-generated answers and automated research task management.
      9
      MIT
    • A
      license
      A
      quality
      C
      maintenance
      An open-source MCP server providing AI agents with neural web search via Exa and tiered web fetch (Exa, local browser, Firecrawl) as a drop-in replacement for built-in web tools, preserving provenance and guarding against SSRF.
      6
      37 PyPI
      9
      MIT