mcp-zotero
Enables interaction with Zotero libraries for searching, managing collections, items, tags, and attachments, plus optional semantic search across PDFs via local embeddings.
README
mcp-zotero
MCP server for Zotero — 38 library tools + optional semantic search via local embeddings.
Install
Base (Zotero API tools only)
uv tool install mcp-zotero
With semantic search
uv tool install "mcp-zotero[rag]"
Adds 8 RAG tools for indexing PDFs and searching by meaning. Downloads a ~80MB embedding model on first use.
With OCR for scanned PDFs
uv tool install "mcp-zotero[rag,ocr]"
Adds docling-based OCR fallback for PDFs that contain scanned images instead of text.
Configuration
Add to your Claude Code MCP settings (.mcp.json or ~/.claude/settings.json):
{
"mcpServers": {
"zotero": {
"command": "mcp-zotero",
"env": {
"ZOTERO_LIBRARY_ID": "YOUR_LIBRARY_ID",
"ZOTERO_LOCAL": "true",
"ZOTERO_LOCAL_KEY": "YOUR_LOCAL_KEY",
"ZOTERO_API_KEY": "YOUR_API_KEY",
"ZOTERO_ATTACHMENTS_DIR": "~/Zotero/storage"
}
}
}
}
Environment variables
| Variable | Required | Default | Description |
|---|---|---|---|
ZOTERO_LIBRARY_ID |
Yes | — | Your Zotero library ID |
ZOTERO_LIBRARY_TYPE |
No | user |
user or group |
ZOTERO_LOCAL |
No | true |
Use Zotero Local API (requires Zotero app running) |
ZOTERO_LOCAL_KEY |
For reads | — | Local API key |
ZOTERO_API_KEY |
For writes | — | Web API key from zotero.org/settings/keys |
ZOTERO_ATTACHMENTS_DIR |
For RAG | — | Path to Zotero storage directory |
RAG-only variables (when installed with [rag])
| Variable | Default | Description |
|---|---|---|
RAG_INDEX_DIR |
~/.zotero-rag |
Where to store index data |
RAG_EMBEDDING_MODEL |
all-MiniLM-L6-v2 |
Sentence transformer model |
RAG_CHUNK_SIZE |
512 |
Target chunk size in tokens |
RAG_CHUNK_OVERLAP |
50 |
Overlap between chunks |
RAG_ENABLE_OCR |
false |
Enable docling OCR fallback for scanned PDFs |
Tools
Base tools (38)
Always available. Covers the full Zotero API:
Search & browse: search_items, top_items, get_item, trash_items, deleted_items
Collections: list_collections, list_collections_top, list_collections_sub, get_collection, collection_items, collection_items_top, collection_tags, create_collections, update_collection, delete_collection, add_item_to_collection, remove_item_from_collection
Tags: list_tags, item_tags, delete_tags
Items: add_item, update_item, delete_item
Attachments: download_attachments, attach_file, attach_linked_file
Schema: item_types, item_fields, item_type_fields, creator_fields, item_creator_types, item_attachment_link_modes
Saved searches: list_searches, list_groups, create_saved_search, delete_saved_search
System: ping, health_check
RAG tools (8)
Available when installed with [rag]. Semantic search across your PDF library:
| Tool | Description |
|---|---|
index_library |
Index a collection or entire library |
index_items |
Index specific items by key |
remove_from_index |
Remove items from the index |
semantic_search |
Search passages by meaning |
get_chunk_context |
Get surrounding text for a search result |
find_similar_chunks |
Find similar passages across documents |
index_status |
Get index statistics |
list_indexed_items |
List all indexed items |
Development
git clone https://github.com/nealcaren/mcp-zotero.git
cd mcp-zotero
uv venv && uv pip install -e ".[dev,rag]"
pytest
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。