Upstage MCP Server

Name: Upstage MCP Server
Author: PritamPatil2603

@PritamPatil2603

リポジトリ

Upstage MCP Server について

A Model Context Protocol server for parsing documents using Upstage AI's document digitization API.

基本情報

カテゴリ

その他

ライセンス

MIT

ランタイム

python

トランスポート

stdio

公開者

PritamPatil2603

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "upstage-mcp-server": {
      "command": "uv",
      "args": [
        "pip",
        "install",
        "upstage-mcp-server"
      ]
    }
  }
}

ツール

** Processes documents and extracts content while preserving structure.

** Extracts structured information from documents based on predefined or auto-generated schemas.

概要

What is Upstage MCP Server?

The Upstage MCP Server is a Model Context Protocol (MCP) server that bridges AI assistants with Upstage AI’s document processing APIs, enabling content extraction and structuring from PDFs, images, and Office files. It is designed for developers and users who want AI models like Claude to parse and extract information from documents.

How to use Upstage MCP Server?

Install using uv (recommended) or pip, then configure Claude Desktop by adding the server entry to claude_desktop_config.json with your UPSTAGE_API_KEY. After restarting Claude, use the parse_document tool to extract structured content or extract_information to retrieve specific data points from a document.

Key features of Upstage MCP Server

Document digitization preserving original layout
Intelligent information extraction with custom schemas
Supports JPEG, PNG, BMP, PDF, TIFF, HEIC, DOCX, PPTX, XLSX
Seamless integration with Claude Desktop and other MCP clients
Auto-generation of extraction schemas

Use cases of Upstage MCP Server

Parse contracts and legal documents for summaries
Extract invoice numbers, dates, and totals from scanned PDFs
Digitize and structure content from multi-page documents
Automate data entry by extracting information from forms

FAQ from Upstage MCP Server

What are the prerequisites to run Upstage MCP Server?

You need an Upstage API key, Python 3.10 or higher, and Astral UV installed.

Where are processing results stored?

Results are saved locally under ~/.upstage-mcp-server/outputs/ in separate folders for document parsing, information extraction, and generated schemas.

What file size and page limits does the server support?

Documents must be under 50 MB and contain fewer than 100 pages to process successfully.

How can I troubleshoot server connection issues?

Ensure uvx is installed and accessible in your PATH, or use its absolute path in the configuration. Restart Claude Desktop and check log files located in %APPDATA%\Claude\logs\ (Windows) or ~/Library/Logs/Claude/ (macOS).

What authentication method does the server use?

It requires an UPSTAGE_API_KEY set in the environment variables within the claude_desktop_config.json file. Obtain a valid key from the Upstage Console.