/ 目錄 / 演練場 / Unstructured UNS MCP
● 官方 unstructured-io ⚡ 即開即用

Unstructured UNS MCP

作者 unstructured-io · unstructured-io/uns-mcp

Parse, chunk, and embed any document with Unstructured's pipeline.

Unstructured's library extracts clean text and structure from messy documents (PDFs, scans, PPTX, emails). This MCP wraps the Unstructured Serverless pipeline so Claude can ingest a folder of raw files and turn them into a queryable RAG corpus.

為什麼要用

核心特性

即時演示

實際使用效果

unstructured-uns-mcp.replay ▶ 就緒
0/0

安裝

選擇你的客戶端

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "unstructured-uns-mcp": {
      "command": "uvx",
      "args": [
        "uns-mcp"
      ]
    }
  }
}

開啟 Claude Desktop → Settings → Developer → Edit Config。儲存後重啟應用。

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "unstructured-uns-mcp": {
      "command": "uvx",
      "args": [
        "uns-mcp"
      ]
    }
  }
}

Cursor 使用與 Claude Desktop 相同的 mcpServers 格式。專案級設定優先於全域。

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "unstructured-uns-mcp": {
      "command": "uvx",
      "args": [
        "uns-mcp"
      ]
    }
  }
}

點擊 Cline 側欄中的 MCP Servers 圖示,然後選 "Edit Configuration"。

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "unstructured-uns-mcp": {
      "command": "uvx",
      "args": [
        "uns-mcp"
      ]
    }
  }
}

格式與 Claude Desktop 相同。重啟 Windsurf 生效。

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "unstructured-uns-mcp",
      "command": "uvx",
      "args": [
        "uns-mcp"
      ]
    }
  ]
}

Continue 使用伺服器物件陣列,而非映射。

~/.config/zed/settings.json
{
  "context_servers": {
    "unstructured-uns-mcp": {
      "command": {
        "path": "uvx",
        "args": [
          "uns-mcp"
        ]
      }
    }
  }
}

加入 context_servers。Zed 儲存後熱重載。

claude mcp add unstructured-uns-mcp -- uvx uns-mcp

一行命令搞定。用 claude mcp list 驗證,claude mcp remove 移除。

使用場景

實戰用法: Unstructured UNS MCP

Stand up a RAG pipeline from a SharePoint folder to Pinecone

👤 Data engineers / RAG builders ⏱ ~15 min intermediate

何時使用: You have a corporate doc dump and want a Claude-queryable index.

前置條件
  • Server/skill installed and authenticated — See repo README
步驟
  1. Define the pipeline
    Create an Unstructured pipeline: source SharePoint folder X, partition by_title with 1024 token max, embed with text-embedding-3-small, target Pinecone index 'corp-docs'.✓ 已複製
    → Pipeline id
  2. Run and monitor
    Run it and tell me when it's done. Report any failed documents.✓ 已複製
    → Status updates + final summary

結果: Production-grade ingest with proper chunking — not naive PDF text dumps.

注意事項
  • Default chunkers can split tables across chunks. For dense tabular docs, use 'by_title' with combine_text_under_n_chars. — Default chunkers can split tables across chunks. For dense tabular docs, use 'by_title' with combine_text_under_n_chars.
搭配使用: filesystem · qdrant-mcp-server

組合

與其他 MCP 搭配,撬動十倍槓桿

unstructured-uns-mcp + filesystem

Pair with filesystem for complementary capabilities

Use this server together with filesystem to complete a multi-step task.✓ 已複製
unstructured-uns-mcp + qdrant-mcp-server

Pair with qdrant-mcp-server for complementary capabilities

Use this server together with qdrant-mcp-server to complete a multi-step task.✓ 已複製

工具

此 MCP 暴露的能力

工具輸入參數何時呼叫成本
create_pipeline source, dest, partition_args, chunk_args Define a new ingest job 1 API call
run_pipeline pipeline_id Execute the pipeline Per Unstructured plan
list_workflows (none) See all configured workflows 1 API call

成本與限制

運行它的成本

API 配額
See provider docs for rate limits
每次呼叫 Token 數
Varies by tool
費用
See repo README for pricing details
提示
Cache tool results and avoid repeated identical calls.

安全

權限、密鑰、影響範圍

憑證儲存: Use environment variables; never commit secrets
資料出站: Tool calls go to the provider's API as documented

故障排查

常見錯誤與修復

401 Unauthorized

Get an API key at unstructured.io → Settings → API keys; set UNSTRUCTURED_API_KEY and UNSTRUCTURED_API_URL.

驗證: list_workflows returns at least one
source connector auth failure

Connector creds (e.g. SharePoint app token) are configured per-source. Recreate the source via Unstructured UI to refresh.

驗證: Run a 1-file test pipeline first

替代方案

Unstructured UNS MCP 對比其他方案

替代方案何時用它替代權衡
LlamaIndex / llamacloudYou want a managed RAG-as-a-service productHigher-level but less control over chunking

更多

資源

📖 閱讀 GitHub 上的官方 README

🐙 查看未解決的 issue

🔍 瀏覽全部 400+ MCP 伺服器和 Skills