Skip to main content

Local MCP servers (Stdio)

MCP servers enable powerful local integrations that work only in the desktop application. The web version cannot use stdio-based MCP servers due to browser security limitations.

Desktop vs. Web comparison​

FeatureDesktop AppWeb App
Stdio MCP✅ Full support❌ Not available
SSE MCP✅ Supported✅ Supported
Local tools✅ Execute locally✅ Limited
File system✅ Full access❌ Limited
Command line✅ Direct access❌ Not available

Why? Browsers cannot execute command-line programs for security reasons. The desktop app runs directly on your computer, so it can use powerful local tools via stdio MCP servers.

What is MCP?​

The Model Context Protocol (MCP) is a standard for connecting AI assistants to tools and data sources. MCP servers:

  • ✅ Run locally on your computer
  • ✅ Communicate via standard input/output (stdio)
  • ✅ Provide tools and data to the AI
  • ✅ Keep everything on your computer
  • ✅ Completely free and open source

How it works​

You → Desktop App → MCP Server → Tool/Data
        (chat)                 (execution)

The AI uses available MCP servers when you ask it to perform tasks that require tools.

Finding MCP servers​

Official directory​

Browse available MCP servers at:

  • Productivity: Task automation, note-taking, file management
  • Development: Git, databases, code tools
  • Data: Analytics, research tools, data processing
  • Education: Learning tools, tutoring systems, content creation

For students, teachers, and educational institutions, these servers are particularly valuable:

Research & Writing​

  • arxiv - Access research papers
  • web-search - Search the internet
  • wikipedia - Look up information
  • exa - Advanced search capabilities

Code & Learning​

  • github - Access repositories
  • bash - Execute shell commands (with caution)
  • python - Run Python code
  • repl - Interactive programming

School Tasks​

  • file-system - Work with local files
  • calculator - Mathematical operations
  • weather - Weather information
  • calendar - Schedule management

How to register an MCP server​

Step 1: Find and install​

  1. Choose a server from the directory
  2. Follow the installation instructions
  3. Note the command to start the server

Step 2: Configure in desktop app​

  1. Open AI-School Desktop
  2. Go to Settings → MCP Servers
  3. Click Add Server
  4. Enter:
    • Name: Server name
    • Command: How to start the server
    • Arguments: Any arguments needed
  5. Click Save

Step 3: Use in chat​

Once registered, the server's tools are automatically available:

You: Use the web-search tool to find information about quantum computing
AI: [uses the tool and provides results]

Example installations​

Web Search Server​

# Install
npm install -g @server/web-search

# Start
web-search

GitHub Server​

# Install
brew install github-mcp  # or your package manager

# Start with token
GITHUB_TOKEN=your_token github-mcp

Python REPL Server​

# Install
pip install mcp-python

# Start
mcp-python

Building your own MCP server​

You can create custom MCP servers for your specific needs!

Why build one?​

  • Automate repetitive tasks
  • Integrate custom tools
  • Connect to proprietary systems
  • Share with your team

Getting started​

  1. Visit MCP Protocol Documentation
  2. Choose your language (Python, JavaScript, Go, etc.)
  3. Follow the tutorial
  4. Deploy to your desktop app

Simple Python example​

from mcp.server import Server
from mcp.tools import Tool

server = Server("my-tool")

@server.tool
async def calculate(expression: str) -> str:
    return str(eval(expression))

server.run()

Register it in the desktop app and start using it!

Safety and security​

Run with confidence​

  • ✅ MCP servers run locally on your computer
  • ✅ No data leaves your computer
  • ✅ You control which servers are active
  • ✅ Disable servers you don't need

Best practices​

  1. Install from trusted sources - Use official repositories
  2. Review the code - Check open-source implementations
  3. Test first - Try in a test folder before important tasks
  4. Disable unused servers - Remove servers you're not using
  5. Keep updated - Install security updates

Troubleshooting​

Server not appearing in the list?

  • Verify the server is installed correctly
  • Check the start command is correct
  • Restart the desktop application

"Server failed to start" error?

  • Check server installation
  • Verify any required environment variables
  • See server documentation for details

Tools not available in chat?

  • Confirm server is running
  • Reload the chat conversation
  • Check server status in settings