egnyte-mcp-server
The Egnyte MCP Server connects to Egnyte domains offering document search and retrieval for AI agents using Egnyteβs API and Python SDK. It supports secure access to enhance AI-driven workflows while respecting user permissions.
Egnyte MCP Server
An MCP (Model Context Protocol) server that connects to your Egnyte domain and exposes document search and retrieval functionality for use in AI agents like Cursor, Claude, or OpenAI.
This server uses Egnyteβs public APIs and Python SDK to support secure, real-time access to your Egnyte content, enabling smart AI-driven workflows while respecting existing permissions.
π About
Model Context Protocol (MCP) is a framework to help AI agents securely query external systems for real-time context.
The Egnyte MCP Server allows agents to:
- Search for documents by name
- Retrieve relevant documents from Egnyte
- Seamlessly integrate enterprise content into generative AI workflows
π οΈ Tools Implemented
Tool Name | Description |
---|---|
search_for_document_by_name | Searches for a document in your Egnyte domain using its filename. |
π Requirements
- Python 3.11+
- Egnyte API access token - Register on https://developers.egnyte.com/member/register to get API key for your Egnyte account
- An Egnyte domain with files to test
Installing Prerequisites
π§ Installation
1. Clone the Repository
git clone https://github.com/egnyte/egnyte-mcp-server.git
cd egnyte-mcp-server
2. Install uv
(Python environment & dependency manager)
Mac/Linux:
curl -LsSf https://astral.sh/uv/install.sh | sh
Windows (PowerShell):
irm https://astral.sh/uv/install.ps1 | iex
3. Install Egnyte SDK
uv pip install egnyte
π Egnyte SDK Documentation
π Egnyte SDK GitHub
4. Setting up Environment Variables
- Create a
.env
File
Create a .env
file inside the root directory with the following content:
DOMAIN=your-egnyte-domain.egnyte.com
ACCESS_TOKEN=your-access-token-here
- Update with Your Credentials
-
Replace
your-egnyte-domain.egnyte.com
with your actual Egnyte domain. -
Replace
your-access-token-here
with your actual Egnyte API access token.This
.env
file is necessary for the server to authenticate and connect securely to your Egnyte domain.
π Running the MCP Server
uv run server.py --python 3.11
This will start the MCP server locally and make the tools available to MCP-compliant clients.
β‘ Setting up MCP Clients
This is a sample MCP (Model Context Protocol) client that connects to a locally running MCP server using fastmcp
.
It uses Python Stdio Transport to communicate with the server and call specific tools by name.
How it works
- Connects to the MCP server (
server.py
) via Python Stdio. - Lists available tools exposed by the server.
- Calls a specific tool by its name, with provided arguments.
- The response is returned based on the tool execution.
Prerequisites
- Python 3.11+
fastmcp
library installed:uv pip install fastmcp
- MCP server (
server.py
) running locally .env
file configured
Usage
- Ensure the MCP server is running first.
- Run the client using the command:
python client.py
Cursor IDE Setup
- Open Cursor β Settings β MCP
- Click "Add new global MCP server"
- Add the following configuration:
{
"mcpServers": {
"Egnyte Document Retriever": {
"command": "uv",
"args": [
"--directory",
"/path/to/egnyte-mcp-server",
"run",
"server.py"
]
}
}
}
β
Replace /path/to/egnyte-mcp-server
with your actual directory path.
- Save and enable the server in the MCP settings.
πΌοΈ Example Screenshots
Cursor MCP Server Configuration
Cursor MCP Query in Action
Claude Desktop Setup
To connect Egnyteβs MCP server to Claude Desktop:
- Launch Claude Desktop and open MCP tool configuration
- Go to Settings β MCP Tools β Add New Server
- Add the following configuration:
{
"mcpServers": {
"egnyte": {
"command": "python3",
"args": ["server.py"],
"cwd": "/Users/yourname/path/to/egnyte-mcp-server",
"env": {
"DOMAIN": "your-egnyte-domain.egnyte.com",
"ACCESS_TOKEN": "your-access-token"
}
}
}
}
β
Replace /Users/yourname/path/to/egnyte-mcp-server
with your actual directory path.
- Save and Start - Claude should detect the server, list search_for_document_by_name under βAvailable MCP tools,β and be able to call it with filenames.
πΌοΈ Example Screenshots
Claude MCP Server Configuration
Claude MCP Query in Action
π Roadmap
- β File search by name
- β Retrieve relevant documents
- π Hybrid search (structured + content)
- π Document text extraction
#Helping with development
Please report any problems you find to or
If you'd like to fix something yourself, please fork this repository, commit the fixes and updates to tests, then set up a pull request with information what you're fixing.