Skip to main content
Glama

Suppr MCP - Usage Guide | 提供文档翻译及中文搜Pubmed的MCP服务 | Suppr超能文献

Suppr MCP Server

Suppr (超能文献) is an AI-powered academic tools platform by WildData. This MCP server brings document translation and literature search capabilities to AI assistants.

🌐 AI Document Translation — Translate PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx), TXT, and HTML documents across 13 languages. Preserves original formatting. Automatic source language detection.

🔬 PubMed Academic Search — Semantic literature discovery across millions of biomedical research papers. Returns structured metadata: DOI, PMID, journal impact factor, citation counts, author affiliations, abstracts, and direct paper links.

🤖 MCP Compatible — Works with Claude Desktop, Cursor, Windsurf, and any Model Context Protocol client.

npm version License: MIT

Install

```bash npx suppr-mcp ```

Related MCP server: Paperlib MCP

Quick Start

1. 安装

全局安装:

npm install -g suppr-mcp

或者使用 npx(无需安装):

npx suppr-mcp

2. 获取 API Key

访问 Suppr API 获取您的 API 密钥。

3. 配置环境变量

export SUPPR_API_KEY=your_api_key_here

4. 在 MCP 客户端中使用

Claude Desktop 配置

编辑 ~/Library/Application Support/Claude/claude_desktop_config.json(macOS)或相应配置文件:

{
  "mcpServers": {
    "suppr": {
      "command": "npx",
      "args": ["-y", "suppr-mcp"],
      "env": {
        "SUPPR_API_KEY": "your_api_key_here"
      }
    }
  }
}

或者使用全局安装:

{
  "mcpServers": {
    "suppr": {
      "command": "suppr-mcp",
      "env": {
        "SUPPR_API_KEY": "your_api_key_here"
      }
    }
  }
}

可用工具

1. create_translation - 创建翻译任务

创建文档翻译任务。

参数:

  • file_path (file_path 和 file_url 二选一): 源文件路径

  • file_url (file_path 和 file_url 二选一): 要翻译的文档 URL

  • to_lang (必填): 目标语言代码

  • from_lang (可选): 源语言代码(默认自动检测)

  • optimize_math_formula (可选): 优化数学公式(仅 PDF)

示例:

{
  "file_url": "https://example.com/document.pdf",
  "to_lang": "en",
  "from_lang": "zh",
  "optimize_math_formula": true
}

返回:

{
  "task_id": "02a6c6d1-3f70-4a5a-80bc-971d53a37bb1",
  "status": "INIT",
  "consumed_point": 453,
  "source_lang": "zh",
  "target_lang": "en",
  "optimize_math_formula": true
}

2. get_translation - 获取翻译详情

获取翻译任务的详细信息和状态。

参数:

  • task_id (必填): 翻译任务 ID

示例:

{
  "task_id": "02a6c6d1-3f70-4a5a-80bc-971d53a37bb1"
}

返回:

{
  "task_id": "02a6c6d1-3f70-4a5a-80bc-971d53a37bb1",
  "status": "DONE",
  "progress": 1.0,
  "consumed_point": 453,
  "source_file_name": "document.pdf",
  "source_file_url": "https://example.com/source.pdf",
  "target_file_url": "https://example.com/translated.pdf",
  "source_lang": "zh",
  "target_lang": "en",
  "error_msg": null,
  "optimize_math_formula": true
}

任务状态说明:

  • INIT: 初始化

  • PROGRESS: 进行中

  • DONE: 已完成

  • ERROR: 错误

3. list_translations - 列出翻译任务

获取翻译任务列表,支持分页。

参数:

  • offset (可选): 分页偏移量,默认 0

  • limit (可选): 每页数量,默认 20

示例:

{
  "offset": 0,
  "limit": 10
}

返回:

{
  "total": 42,
  "offset": 0,
  "limit": 10,
  "list": [
    {
      "task_id": "...",
      "status": "DONE",
      "progress": 1.0,
      ...
    }
  ]
}

4. search_documents - 文献搜索

AI 驱动的文献语义搜索。

参数:

  • query (必填): 自然语言查询

  • topk (可选): 最大返回数量(1-100,默认 20)

  • return_doc_keys (可选): 指定返回字段

  • auto_select (可选): 自动选择最优结果(默认 true)

示例:

{
  "query": "糖尿病最新研究进展",
  "topk": 5,
  "return_doc_keys": ["title", "abstract", "doi", "authors"],
  "auto_select": true
}

可用的返回字段:

  • title: 标题

  • abstract: 摘要

  • authors: 作者列表

  • doi: DOI

  • pmid: PubMed ID

  • link: 链接

  • publication: 出版物

  • pub_year: 出版年份

  • 更多字段请参考 API 文档

返回:

{
  "search_items": [
    {
      "doc": {
        "title": "...",
        "abstract": "...",
        "authors": [...],
        "doi": "...",
        ...
      },
      "search_gateway": "pubmed"
    }
  ],
  "consumed_points": 20
}

支持的语言

常用语言代码:

  • en: English (英语)

  • zh: Chinese (中文)

  • ko: Korean (韩语)

  • ja: Japanese (日语)

  • fr: French (法语)

  • de: German (德语)

  • es: Spanish (西班牙语)

  • ru: Russian (俄语)

  • ar: Arabic (阿拉伯语)

  • pt: Portuguese (葡萄牙语)

  • it: Italian (意大利语)

  • auto: 自动检测

错误处理

所有错误都会返回标准格式:

{
  "code": 非零错误码,
  "msg": "错误信息",
  "data": null
}

常见错误:

  • 401: API 密钥无效或未提供

  • 400: 请求参数错误

  • 404: 资源不存在

使用示例

在 Claude Desktop 中使用

  1. 配置好 API 密钥后重启 Claude Desktop

  2. 在对话中使用工具:

翻译文档:

请帮我翻译这个文档:https://example.com/paper.pdf,翻译成英文

搜索文献:

帮我搜索关于"深度学习在医学影像中的应用"的最新文献

查询翻译状态:

查看任务 02a6c6d1-3f70-4a5a-80bc-971d53a37bb1 的翻译进度

常见问题

Q: 如何获取 API 密钥?

A: 访问 https://suppr.wilddata.cn/api-keys 注册并获取 API 密钥。

Q: 支持哪些文档格式?

A: 支持 PDF, DOCX, PPTX, XLSX, HTML, TXT, EPUB等常见格式。

Q: 翻译需要多长时间?

A: 取决于文档大小,通常几分钟到十几分钟不等。可以使用 get_translation 查询进度。

Q: 如何下载翻译后的文档?

A: 翻译完成后,get_translation 会返回 target_file_url,直接访问该链接下载。

Q: npx 运行失败?

A: 确保 Node.js 版本 >= 18.0.0,并且设置了 SUPPR_API_KEY 环境变量。

🔗 Suppr超能文献产品

技术支持

如需帮助,请联系:IT@wilddata.cn

Made with ❤️ by WildData

Suppr Ecosystem

Product

Link

🌐 Suppr Platform

suppr.wilddata.cn

📖 API Documentation

openapi.suppr.wilddata.cn

🔌 Zotero Plugin

suppr-zotero-plugin

🤖 Claude Code Skills

suppr-skills

🔬 Deep Research

suppr.wilddata.cn/deep-research

📄 AI Translation

suppr.wilddata.cn/translate/upload

🔎 PubMed Search

suppr.wilddata.cn

Available Tools

4 tools
create_translationCreate Translation TaskB

Create a document translation task. Supports file upload via path or URL.

Inputs: to_lang (required), file_url, file_path, from_lang, +1 more.

get_translationGet Translation StatusA

Get translation task details and status. Use this to check progress and get result URLs.

Inputs: task_id (required).

list_translationsList Translation TasksA

List translation tasks with pagination. View all historical translation tasks.

Inputs: limit, offset.

search_documentsSearch Academic LiteratureB

AI-driven semantic search for academic literature. Input natural language query to find relevant papers.

Inputs: query (required), topk, auto_select, return_doc_keys.

Tool Schema Changelog

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

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 4 tools

    Disambiguation5/5

    Each tool has a clear, distinct purpose: create, get, and list translations, plus search documents. There is no overlap or ambiguity between them.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern: create_translation, get_translation, list_translations, and search_documents. No mixed conventions.

    Tool Count4/5

    Four tools is a reasonable size for a server handling translation tasks and document search. It feels slightly minimal but each tool serves a distinct purpose.

    Completeness3/5

    The translation lifecycle covers create, get, and list, but lacks update/delete/cancel operations. The search_documents tool seems unrelated to translations, creating a mixed domain with notable gaps.

    Maintenance

    ActivityInactive
    ResponsivenessLimited data

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables comprehensive biomedical literature research through PubMed database access with advanced search, full-text retrieval, citation analysis, and batch processing capabilities. Supports both local deployment and cloud hosting for seamless integration with AI assistants.
      1
      MIT
    • F
      license
      C
      quality
      B
      maintenance
      Enables academic literature management through PDF import, hybrid search, knowledge graph construction, and automated literature review generation. Combines full-text search with semantic vector search for comprehensive paper analysis.
      55
      -
    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables PDF document processing including text, image, and table extraction, as well as intelligent classification and similarity analysis across multiple languages.
      51
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables uploading, organizing, and semantically searching documents with support for various file types and embedding providers.
      28 npm
      MIT