Skip to main content

​
Summary

Start and connect to a local Stdio Model Context Protocol (MCP) server. This extends AI model capabilities with custom tools and functions provided by MCP servers.
  • Needs Admin: False
  • Version: 0
  • Author: @Ne0nd0g

​
Arguments

​
command

  • Description: The command or program to start the MCP server
  • Required: True
  • Type: String

​
args

  • Description: Arguments to pass to the command
  • Required: False
  • Type: Array of Strings

​
Usage

mcp-connect -command uv -args --directory /opt/mythic_mcp/ run main.py mythic_admin password 127.0.0.1 7443
mcp-connect -command python3 -args /path/to/mcp_server.py
mcp-connect -command node -args /path/to/mcp-server.js

​
Detailed Summary

The mcp-connect command launches a Stdio-based Model Context Protocol server and establishes a connection to it. MCP servers provide additional tools and capabilities that AI models can use during inference.

​
Model Context Protocol (MCP)

MCP is a standardized protocol for connecting AI models to external tools and data sources. When an MCP server is connected:
  • Models can discover available tools
  • Tools can be invoked during chat or query operations
  • Results are seamlessly integrated into responses
  • Enhanced reasoning and action capabilities are available

​
Stdio MCP Requirements

Stdio MCP servers must run in the same location as the Sage container. All dependencies (Python, Node.js, etc.) must be pre-installed.
Container Deployment If Sage runs in Docker:
  • Install MCP server dependencies in the Sage container
  • Modify the Dockerfile to include required packages
  • Place MCP server code inside the container
  • Use container-internal paths in the command
Local Deployment If Sage runs directly on the Mythic host:
  • Install MCP server dependencies on the host
  • Use host filesystem paths
  • Ensure proper permissions for the Sage process

​
Mythic MCP Server

Sage includes the Mythic MCP server pre-installed at /opt/mythic_mcp/. Connection Example 
mcp-connect -command uv -args --directory /opt/mythic_mcp/ run main.py mythic_admin SuperSecretPassword 127.0.0.1 7443
Arguments Breakdown
  • uv - Command to run (UV Python package runner)
  • --directory - Working directory for the MCP server
  • /opt/mythic_mcp/ - Location of Mythic MCP code
  • run - UV subcommand to execute the server
  • main.py - MCP server entry point
  • mythic_admin - Mythic username for API access
  • SuperSecretPassword - Mythic user password
  • 127.0.0.1 - Mythic server IP (not loopback if containerized)
  • 7443 - Mythic server port
When running in a container, use the actual IP address instead of 127.0.0.1 or localhost for Mythic connectivity.

​
Custom MCP Servers

You can connect to any Stdio-based MCP server: Python MCP Server 
mcp-connect -command python3 -args /path/to/server.py
Node.js MCP Server 
mcp-connect -command node -args /path/to/server.js --config /path/to/config.json
Compiled Binary 
mcp-connect -command /opt/custom-mcp/mcp-server -args --verbose --port 8080

​
Tool Integration

Once connected, MCP tools become available to models: In Chat Sessions
  • Tools are automatically discovered
  • Models can request tool usage
  • Results are incorporated into conversation
In Query Operations
  • Single queries can use MCP tools
  • Tool results appear in the response
  • Enable with the tools parameter

​
Troubleshooting

Connection Failures
  • Verify command path is correct
  • Check all dependencies are installed
  • Ensure arguments are properly formatted
  • Review MCP server logs for errors
Permission Issues
  • Confirm Sage process has execution permissions
  • Check file paths are accessible
  • Verify network access if MCP calls external services
Tool Discovery Problems
  • Ensure MCP server implements the protocol correctly
  • Check server initialization completes successfully
  • Verify tools are properly registered
MCP servers run as child processes of Sage. If Sage restarts, MCP connections are lost and must be re-established.

​
Security Considerations

  • MCP servers have access to Sage’s execution context
  • Validate and trust MCP server code before connecting
  • Limit MCP server permissions when possible
  • Audit tool calls in production environments
  • Consider network isolation for sensitive operations

​
Best Practices

  • Test MCP connections in a development environment first
  • Document required dependencies for your MCP servers
  • Use environment variables for sensitive configuration
  • Monitor MCP server resource usage
  • Implement proper error handling in custom MCP servers