spinchange/umi
Universal Machine Interface — An open schema for how AI agents query and understand machines.
Platform-specific configuration:
{
"mcpServers": {
"umi": {
"command": "npx",
"args": [
"-y",
"umi"
]
}
}
}Add the config above to .claude/settings.json under the mcpServers key.
[](https://github.com/spinchange/umi/actions/workflows/test.yml) [](https://codecov.io/gh/spinchange/umi) [](LICENSE) [](https://pypi.org/project/umi-mcp/) []() [](https://modelcontextprotocol.io)
A Python MCP server that gives AI assistants live awareness of the machine they're running on.
Ask your AI assistant how much disk space is left, what's using the most memory, or whether any services have crashed recently — and get a real answer, not a suggested command to run yourself.
---
UMI runs as a local Model Context Protocol server. Once installed, compatible AI clients can call its tools to query your system and reason about what they find.
You: "Is anything on this machine worth worrying about?"
AI: "C: is at 75% capacity, RAM is 87% used (866 MB free of 8 GB),
and the Claude Desktop service has crashed 3 times in the last
week according to the event log. Everything else looks normal."That response required no commands, no copy-pasting, and no prompting the user to open Task Manager. The AI queried the machine directly, assembled the picture, and answered in plain language.
---
| Tool | What it returns | |------|----------------| | get_umi_uptime | Hostname, OS, architecture, boot time, uptime, CPU count, RAM, current CPU and memory utilization, swap usage, load averages | | get_umi_disk | Mounted volumes — capacity, used/free, filesystem, I/O counters (reads, writes, bytes, time) | | get_umi_network | Network i
Loading reviews...