ENL-Reader-MCP
This project provides an automated service for reading EndNote `.enl` libraries and exposing their contents via the MCP (Modular Command Protocol) interface. It enables structured, programmatic access to bibliographic data and PDF full texts for downstream LLMs or automation clients.
Documentation
EndNote MCP Service
This project provides an automated service for reading EndNote `.enl` libraries and exposing their contents via the MCP (Modular Command Protocol) interface. It enables structured, programmatic access to bibliographic data and PDF full texts for downstream LLMs or automation clients.
Features
- Automatic EndNote Library Parsing: Reads `.enl` (SQLite) files and associated `.Data` folders.
- MCP Tool Interface: Exposes all functions as MCP tools for easy integration with LLMs or workflow engines.
- List All References: Enumerate all bibliographic entries with metadata.
- Fuzzy Search: Search references by (partial) title or keywords, supporting both English and Chinese.
- PDF Full Text Extraction: Retrieve and extract the full text of attached PDF files by paper title.
- Backup Mode & Manual Refresh: Optionally operate on a `.enl.backup` file for safer access, with both automatic and manual refresh support.
Typical Usage
- List all papers: `list_papers()`
- Fuzzy search: `search_papers('distillation')`
- Extract full text: `read_paper('Paper Title')`
- Manually refresh backup: `refresh_backup()`
Application Scenarios
- LLM-based literature review and knowledge extraction
- Automated research assistants
- Academic data mining and analysis
Requirements
- Python 3.10+
- EndNote `.enl` file and corresponding `.Data` folder
- MCP/fastmcp runtime environment
Installation
1. Install uv if you don't have it:
pip install uv
# or see https://github.com/astral-sh/uv for other installation methods2. Install dependencies using uv (recommended):
uv pip install -r requirements.txtThe main dependencies are:
A requirements.txt file is provided for compatibility, but uv is the recommended tool for dependency management.
Configuration
You need to provide the path to your EndNote `.enl` file and the corresponding `.Data` folder. Optionally, you can enable detailed logging.
You may also enable backup mode with the `--use-backup` or `-b` option. When enabled, all database operations are performed on a `.enl.backup` file, which is automatically refreshed from the original `.enl` file at startup. You can manually refresh the backup at any time using the `refresh_backup()` MCP tool.
> Why Backup Mode?
>
> EndNote software locks the `.enl` file while it is running, preventing other programs from accessing it. This means that when you are viewing or editing references in EndNote, this tool cannot operate on the original `.enl` file. To solve this, backup mode was introduced: all operations are performed on a `.enl.backup` copy. When you need to refresh the backup, you must first close EndNote to release the file lock; otherwise, the refresh will fail.
Running the Server
Start the MCP server with the required arguments:
python server.py --enl-file --data-folder [--enable-log] [--use-backup]
# or using short options:
python server.py -e -d -l -b- `--enl-file`, `-e`: Path to the EndNote `.enl` file (required)
- `--data-folder`, `-d`: Path to the EndNote `.Data` folder (required)
- `--enable-log`, `-l`: Enable detailed log output (optional)
- `--use-backup`, `-b`: Use `.enl.backup` for all DB operations (optional, enables backup mode; the backup is auto-refreshed at startup)
When backup mode is enabled, all operations are performed on the `.enl.backup` file. The backup is automatically refreshed from the original `.enl` file at server startup, and you can manually refresh it at any time using the `refresh_backup()` tool.
MCP Server Configuration
To integrate this service with MCP-compatible clients or orchestration tools, you can define an entry in your `mcp.json` configuration. Example:
{
"mcpServers": {
"enl_reader_mcp": {
"command": "uv",
"args": [
"run",
"--directory",
"",
"-m",
"server",
"-e",
"E:/EndNoteLib/your_library.enl",
"-d",
"E:/EndNoteLib/your_library.Data",
"-l",
"-b"
]
}
}
}- Replace `` with the actual path where you cloned this repository.
- Adjust the paths to your `.enl` file and `.Data`
- Add `-b` to enable backup mode if desired. When enabled, the service will operate on a `.enl.backup` file and support manual refresh via the `refresh_backup` tool.
MCP Tools
- `list_papers(offset=0, limit=10)`: List all references with pagination.
- `search_papers(query)`: Fuzzy search references by title or keywords.
- `read_paper(title)`: Get metadata and PDF full text by (fuzzy) title.
- `refresh_backup()`: Manually refresh the `.enl.backup` file (only available when backup mode is enabled; has no effect if backup mode is off).
Note: The `refresh_backup` tool is only effective when backup mode is enabled. Backup mode is recommended for scenarios requiring read-only or safer access to the EndNote library.
Frequently asked questions
What is ENL-Reader-MCP?
ENL-Reader-MCP is This project provides an automated service for reading EndNote `.enl` libraries and exposing their contents via the MCP (Modular Command Protocol) interface. It enables structured, programmatic access to bibliographic data and PDF full texts for downstream LLMs or automation clients.
How do I install ENL-Reader-MCP?
Open the GitHub repository and follow its README. Most MCP servers are added to your client's MCP config, then called by your agent.
Is ENL-Reader-MCP open source?
Yes — it is hosted on GitHub at https://github.com/DOGGY-SAINT/ENL-Reader-MCP and has 9 stars.
Related MCP tools
Cognee is the open-source AI memory platform for agents. Give your AI agents persistent long-term memory across sessions with a self-hosted knowledge graph engine.
Automate browser based workflows with AI
Hindsight: Agent Memory That Learns
A privacy-first app that strips AI watermarks from content you own.
Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc.
The power of Claude Code / GeminiCLI / CodexCLI + [Gemini / OpenAI / OpenRouter / Azure / Grok / Ollama / Custom Model / All Of The Above] working as one.
Run your own MCP server? See who uses it and what to fix.
Measure it with TrackMCP