system-mcp
README.md
# system-mcp
A local MCP server (Python + `uv`) that exposes system telemetry tools:
- CPU usage and core/frequency metrics
- RAM and swap usage
- Disk usage and IO
- Battery status
- Network counters
- Running task/process stats
- GPU metrics (via `nvidia-smi` when available)
## 1) Install
```bash
uv sync
```
## 2) Run server (stdio transport)
```bash
uv run system-mcp
```
## 3) MCP client configuration example
Use this command in your MCP client config:
- Command: `uv`
- Args: `run`, `system-mcp`
- Working directory: this project root
Example JSON snippet:
```json
{
"mcpServers": {
"system-mcp": {
"command": "uv",
"args": ["run", "system-mcp"],
"cwd": "k:/Tech/Projects/system_mcp"
}
}
}
```
## Exposed tools
- `system_overview`
- `cpu_metrics(sample_seconds=1.0)`
- `memory_metrics`
- `disk_metrics(path="C:\\")`
- `battery_metrics`
- `network_metrics`
- `running_tasks(limit=10, sort_by="cpu")`
- `gpu_metrics`
## Notes
- `gpu_metrics` depends on `nvidia-smi` being available in `PATH`.
- On systems without battery/GPU, tools return availability flags and messages instead of failing.
TDQS
A3.7/5.0
Scored across 8 tools
Disambiguation5/5
Each tool targets a distinct system component (battery, CPU, disk, GPU, memory, network, processes, and overview) with no functional overlap.
Naming Consistency5/5
All tool names follow a consistent snake_case pattern with descriptive suffixes like _metrics, _tasks, _overview.
Tool Count5/5
8 tools is well-scoped for a system monitoring server, covering major resources without being excessive or too sparse.
Completeness4/5
Core system metrics are covered, but lacks deeper details like per-interface network stats or sensor data, which are minor gaps.
Maintenance
ActivityInactive
ResponsivenessNo issues