sysmcp

sysmcp

MCP server that exposes live host metrics (CPU, memory, processes, network) to Cursor/Claude via /proc and optional eBPF, enabling system monitoring dashboards and natural language queries.

Category
访问服务器

README

sysmcp

MCP server that exposes live host signal — CPU, memory, processes, network — to Cursor / Claude. Default path is /proc (no root). Optional eBPF tools activate when CAP_BPF / CAP_PERFMON is available.

Repo: https://github.com/arunmm8335/sysmcp · Version: 0.5.0

Tools

Tool What it returns
host_dashboard Full snapshot (CPU, mem, top procs, listeners, eBPF)
poll_host_stats Lightweight CPU/mem/process snapshot for auto-refresh
system_stats CPU %, memory %, load averages, NIC rx/tx B/s
top_processes Top PIDs by CPU % over a sample window, plus skipped_count
network_connections IPv4+IPv6 TCP/UDP; filters + resolve_pid; row limit (default 200)
cgroup_cpu Top cgroups by CPU % (cgroup v2 cpu.stat)
iface_rates Per-interface byte/packet rates (/proc/net/dev)
ebpf_status Whether eBPF works; includes how_to_enable when not
ebpf_activity Fork + sched_switch rates via eBPF, or available: false
ebpf_traffic Kernel RX/TX packet rates (netif_receive_skb / net_dev_xmit)

True per-connection byte counters need sock-level eBPF; iface_rates + ebpf_traffic are the practical deeper network signal today.

Browser dashboard (HTTP)

go build -o sysmcp ./cmd/sysmcp
./sysmcp serve
# open http://127.0.0.1:8787/

Optional: ./sysmcp serve -addr 127.0.0.1:9000

APIs: /api/dashboard, /api/poll (JSON).

Interactive views in Cursor

Ask “Open the host dashboard” / “Refresh the host dashboard”. The agent calls host_dashboard and opens a Cursor Canvas beside the chat.

MCP Apps resource ui://sysmcp/dashboard remains for hosts that render Apps iframes. Cursor often uses Canvas + JSON instead.

network_connections filters

Arg Effect
listen_only TCP LISTEN + unbound UDP
protocol tcp / udp
family ipv4 / ipv6
state e.g. ESTABLISHED, LISTEN
local_port filter by local port
pid owning PID (implies resolve_pid)
limit max rows (default 200, max 1000); truncated in result

Install

Release binaries (linux amd64 / arm64) ship on each v* tag:

https://github.com/arunmm8335/sysmcp/releases

# example amd64
curl -fsSL -o sysmcp.tgz \
  https://github.com/arunmm8335/sysmcp/releases/download/v0.5.0/sysmcp_0.5.0_linux_amd64.tar.gz
tar -xzf sysmcp.tgz

Or from source:

go install github.com/arunmm8335/sysmcp/cmd/sysmcp@latest
# or
git clone https://github.com/arunmm8335/sysmcp.git
cd sysmcp
go build -o sysmcp ./cmd/sysmcp

MCP Registry (one-click)

Namespace: io.github.arunmm8335/sysmcp. Tag releases build .mcpb assets and attempt an OIDC publish via mcp-publisher. To publish manually after a release (recommended if OIDC is not set up):

# fill fileSha256 in server.json from the release *.mcpb.sha256 files, then:
curl -fsSL -o mcp-publisher.tgz \
  https://github.com/modelcontextprotocol/registry/releases/download/v1.8.0/mcp-publisher_linux_amd64.tar.gz
tar -xzf mcp-publisher.tgz mcp-publisher
./mcp-publisher login github
./mcp-publisher publish ./server.json

See https://modelcontextprotocol.io/registry/quickstart

Cursor / Claude MCP config

{
  "mcpServers": {
    "sysmcp": {
      "command": "/absolute/path/to/sysmcp"
    }
  }
}

Enable eBPF (optional)

go build -o sysmcp ./cmd/sysmcp
./scripts/grant-caps.sh ./sysmcp
# sets: cap_bpf,cap_perfmon+ep

Restart the MCP server after granting caps. Without caps, eBPF tools return available: false — /proc tools still work.

Demo

go test ./... -count=1
go build -o sysmcp ./cmd/sysmcp
go run ./scripts/smoke.go ./sysmcp
./sysmcp serve   # then open http://127.0.0.1:8787/

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选