Dewey

com.meetdewey由 meetdewey.com 發布官方發布者v1.5.4更新於 Nov 2, 2026

Agentic search over your Dewey document collections from any MCP-compatible client.

已驗證Streamable HTTP可網頁執行檔案與儲存知識與記憶

概覽

AI 產生的概覽

Dewey 讓助理搜尋、閱讀並研究你的 Dewey 文件集合，提供附引用的回答、論點與矛盾分析以及文件管理。

可以做什麼
  • 列出集合并進行語意加關鍵字混合檢索、章節掃描與附引用的代理式研究查詢
  • 以 Markdown 讀取章節與文件，列出文件章節及處理狀態
  • 列出擷取的論點與偵測到的矛盾，觸發矛盾偵測並套用或略過解決方案
  • 執行去重、列出重複群組，並提升或解散標準文件
  • 查看集合統計、更新集合設定與研究指示、重算摘要與圖說，重試或刪除文件與集合
適用情境
當你的研究或參考資料已存放在 Dewey 集合中，並希望助理根據這些內容給出有依據、附引用的回答，或查核文件之間的矛盾與重複時適用。
執行需求
需要 Dewey 專案 API 金鑰（DEWEY_API_KEY）；可選 DEWEY_COLLECTION_ID 作為預設集合，DEWEY_API_URL 覆寫 API 基礎位址。可作為遠端端點使用，或透過 npm 套件 @meetdewey/mcp 搭配 Node.js（npx）在本機執行。去重功能需在集合上啟用 enableDeduplication。
安裝前請注意
API 金鑰是專案憑證，透過環境變數傳入。多個工具會修改或銷毀資料：刪除文件與集合無法復原，更新、重試、重算、矛盾處理與去重執行都會改變集合狀態與指示。文件內容會傳送到 Dewey 的 API 進行處理與研究。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Dewey，將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/com.meetdewey%2Fdewey/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。

{
  "mcpServers": {
    "dewey": {
      "type": "http",
      "url": "https://mcp.meetdewey.com/mcp"
    }
  }
}

README

[Dewey]

@meetdewey/mcp

MCP server for Dewey — search and research your document collections from Claude, Cursor, and any MCP-compatible agent. See the full API reference for details.

Installation

Add to your claude_desktop_config.json (or equivalent MCP client config):

json
{  "mcpServers": {    "dewey": {      "command": "/bin/sh",      "args": ["-lc", "npx -y @meetdewey/mcp"],      "env": {        "DEWEY_API_KEY": "dwy_live_...",        "DEWEY_COLLECTION_ID": "..."      }    }  }}

Note for Windows users: replace "command": "/bin/sh" and "args": ["-lc", "npx -y @meetdewey/mcp"] with "command": "npx" and "args": ["-y", "@meetdewey/mcp"] — the login-shell wrapper is only needed on macOS/Linux to pick up Homebrew and nvm PATH entries that Claude Desktop doesn't inherit.

Environment variables

VariableRequiredDescription
DEWEY_API_KEYYesYour Dewey project API key
DEWEY_COLLECTION_IDNoDefault collection ID. When set, tools that accept collection_id will fall back to this value.
DEWEY_API_URLNoOverride the API base URL (default: https://api.meetdewey.com/v1)

Available tools

Search & research

ToolDescription
dewey_list_collectionsList all collections in the project, including description and research instructions when set
dewey_searchHybrid semantic + keyword search over chunk content
dewey_scan_sectionsLightweight search over section titles and summaries
dewey_researchAgentic research query with configurable depth — returns a grounded, cited answer
dewey_get_sectionFetch the full Markdown content of a section by ID

Document management

ToolDescription
dewey_list_documentsList documents in a collection with their processing status
dewey_get_document_sectionsList all sections in a document (table of contents with heading levels and IDs)
dewey_get_document_markdownFetch the full converted Markdown content of a document
dewey_retry_documentRetry a failed document — clears error state and re-queues processing
dewey_delete_documentPermanently delete a document and all its derived data

Claims & contradictions

ToolDescription
dewey_list_claimsList extracted factual claims from a collection or specific document, filterable by importance (1–5)
dewey_list_contradictionsList detected contradictions — clusters of conflicting claims with explanations and suggested resolutions
dewey_detect_contradictionsTrigger an async contradiction detection run across all claims in a collection
dewey_get_contradiction_runGet the status and stats of the latest contradiction detection run (use to poll after dewey_detect_contradictions)
dewey_resolve_contradictionApply a resolution instruction to a contradiction or dismiss it

Deduplication

ToolDescription
dewey_detect_duplicatesTrigger an async deduplication run — identifies near-duplicate documents by measuring shared content across chunks and marks one per cluster as canonical
dewey_get_duplicate_runGet the status and stats of the latest deduplication run (use to poll after dewey_detect_duplicates)
dewey_list_duplicate_groupsList near-duplicate groups with canonical + members and coverage percentages
dewey_promote_duplicate_canonicalPromote a different member to canonical; old canonical becomes near_duplicate
dewey_disband_duplicate_groupDisband a group; former members rejoin retrieval as distinct documents

Non-canonical documents are excluded from retrieval and contradiction detection. Requires enableDeduplication: true on the collection (set via dewey_update_collection or the Dewey dashboard).

Collection settings

ToolDescription
dewey_get_collection_statsGet document count, storage, section/chunk/claim counts, and processing status breakdown
dewey_update_collectionUpdate collection name, description, research instructions, visibility, and feature flags
dewey_recompute_summariesRe-run AI section summarization across all documents (e.g. after changing the LLM model)
dewey_recompute_captionsRe-run AI captioning for all images and tables across all documents
dewey_delete_collectionPermanently delete a collection and all its data

Research instructions

Collections can have natural-language instructions that are automatically injected into the research system prompt — for example, noting units, preferred sources, or how to handle missing information. They can be set in the Dewey dashboard, via the API, or directly through dewey_update_collection. When dewey_resolve_contradiction is used to apply a resolution, the suggested instruction is also appended here automatically.

Tool usage pattern

A typical read-only research workflow:

  1. dewey_list_collections — discover available collections and understand their purpose
  2. dewey_scan_sections — quickly explore document structure to identify relevant sections
  3. dewey_get_section — load full content of a specific section
  4. dewey_search — retrieve the most relevant chunks for a focused question
  5. dewey_research — run a full agentic research query for questions that need multi-step reasoning

For deeper analysis and curation:

  1. dewey_get_collection_stats — assess how much content has been processed
  2. dewey_list_claims — surface the most important facts across documents
  3. dewey_list_contradictions — identify where documents disagree
  4. dewey_resolve_contradiction — apply or dismiss each contradiction, updating collection instructions automatically

License

MIT

來源：README.md，提交 3d83f44

工具

0
SourceWeft 尚未讀取此伺服器的工具。

版本歷史

1
  1. v1.5.4最新Sep 16, 2026

評分

已滿足適用檢查項的 40%

該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通必需25 分

    請保持端點能夠回應 MCP 握手；SourceWeft 每 30 天探測一次。

  • 已宣告認證方式必需10 分

    請在註冊中心條目中宣告憑證請求標頭，或保持端點能夠回應探測，以便判定其認證方式。

  • 已列出工具15 分

    請讓伺服器在無憑證時回應 tools/list，或發布可在沙箱中無憑證啟動的軟體包。

  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分

    請在儲存庫或註冊中心條目中宣告授權條款。