Zotero MCP Agent
An MCP server that allows an LLM to research online papers, manage your Zotero library, and export knowledge bases.
Installation
- Install Dependencies:
pip install -r requirements.txt(Or use uv sync if using uv)
- Configuration:
- Copy
.env.exampleto.env:
cp .env.example .env- Fill in your keys in
.env(Zotero API Key, LLM Key, Tavily Key). - Proxy: If you are in a restricted network environment (e.g. using v2rayN), ensure
HTTP_PROXYandHTTPS_PROXYare set in.env.
Usage with Chatbox / Claude Desktop
To use this MCP server, you need to configure your client to run the server script.
Command
The server runs via standard input/output (stdio).
Command: python Arguments: [absolute/path/to/main.py]
Example configuration for Claude Desktop claude_desktop_config.json:
{
"mcpServers": {
"zotero-agent": {
"command": "python",
"args": [
"D:/Yan2file/zotero-mcp/main.py"
],
"env": {
"PYTHONPATH": "D:/Yan2file/zotero-mcp"
}
}
}
}
> [!IMPORTANT]
> **Windows Users**: If you encounter errors like `[Errno 2] No such file or directory` where the path looks mashed together:
> 1. Use **forward slashes** (`/`) in your paths.
> 2. Point to `main.py` in the root folder, NOT `src/server.py`.
> Example: `D:/Yan2file/zotero-mcp/main.py`
*Note: You might need to use the full path to your python executable if it's in a virtual environment (e.g. `D:/Yan2file/zotero-mcp/.venv/Scripts/python.exe`).*
## Available Tools
- `search_online_papers(query, download=True)`: Search and download PDFs.
- `read_local_paper(file_path)`: Read the text content of a downloaded PDF.
- `read_zotero_item(item_key)`: Read the content of a Zotero item by its Key (e.g. `ABCD1234`).
- `search_local_library(query)`: Search your Zotero library.
- `export_knowledge_base(output_path)`: Export Zotero items to Markdown.
## Example Prompts
- "Search for the latest papers on LLM Agents and download two of them."
- "Download the paper 'Attention is All You Need' and summarize its key contributions."
