Hey @Amm0, thanks for your feedback.
1. Compiling Python into EXE:
Actually, an MCP page will usually integrate with AI using a JSON file (like claude_desktop_config.json in Claude Desktop), so you can integrate it with your AI by creating a sample JSON file like this:
{
"mcpServers": {
"mikrotik-mcp-server": {
"command": "docker",
"args": [
"run",
"--rm",
"-i",
"-e", "MIKROTIK_HOST=192.168.88.1",
"-e", "MIKROTIK_USERNAME=sshuser",
"-e", "MIKROTIK_PASSWORD=your_password",
"-e", "MIKROTIK_PORT=22",
"mikrotik-mcp"
]
}
}
}
Please check the README to see how you can integrate it with your desktop AI.
2. The results so far:
Here you can find lots of examples of how you can use AI to change your network configuration in your MikroTik. However, lots of tools and functions are NOT SUPPORTED in the mikrotik-mcp yet, but I believe the community will grow and we will support more tools from MikroTik.
Your high-level thought is right, especially with new LLMs like Claude 4.5 Sonnet, AI is doing pretty well in almost all areas. The mikrotik-mcp only provides a way for AI to apply (with user permission) configuration to the MikroTik.
Regarding your feedback about "mapping RouterOS's CLI/API directly into MCP tools may not offer much but IDK" actually, mikrotik-mcp is using SSH connection to the MikroTik, and each tool is separated into different areas. Please check here.
For example, for DNS settings in MikroTik, here is the implementation for the mikrotik-mcp you can find in dns.py. As you can see, we are running direct commands and not only communicating with APIs. So the more tools you add to the folder scope, the more features and capabilities AI will have to integrate with your MikroTik network configuration.
Your assumption is correct, currently we are not covering all of the features, tools, and areas in MikroTik. You may ask your AI to perform a request that needs not only DNS or firewall areas, but also another area which is not supported (YET) in the mikrotik-mcp*. That's why I would like to encourage the community to participate and create issues [here], so the existing developers at mikrotik-mcp, like me, can follow up on the issues and create valid PRs to improve the mikrotik-mcp.*
By combining network engineers' knowledge from this community (by creating issues) and developers (by implementing and addressing the issues), I think we can eventually set up an AI Agent for managing MikroTik devices.
I really believe that soon we will need to enhance AI Agents with MCP to handle any engineering tasks. In the networking area, I believe MikroTik has a lot to offer, so it's good to have its MCP ready for future AI agents!
