能力标签

🔌 MCP 🤖 Agent 🔄 工作流 🐳 Docker 💻 CLI 🔗 REST API 🧠 Claude 🕸 采集

🔌

MCP工具

边缘核心

基于 Elixir · 让 AI 助手直接操作你的系统与工具

英文名：edge-core

⭐ 21 Stars 🍴 1 Forks 💻 Elixir 📄 NOASSERTION 🏷 AI 8.0分

8.0AI 综合评分

mcpautomationedge-aiedge-computing

⚙️ 配置说明 📺 TG 频道

✦ AI Skill Hub 推荐

边缘核心是 AI Skill Hub 本期精选MCP工具之一。综合评分 8.0 分，整体质量较高。我们强烈推荐将其纳入你的 AI 工具库，帮助提升工作效率。

📚 深度解析

边缘核心是一款基于 MCP（Model Context Protocol）标准协议的 AI 工具扩展。MCP 协议由 Anthropic 开发并开源，旨在建立 AI 模型与外部工具之间的标准化通信接口，目前已被 Claude Desktop、Claude Code、Cursor 等主流 AI 工具采纳。

通过安装边缘核心，你的 AI 助手将获得额外的工具调用能力，可以用自然语言直接操控该工具的功能，无需学习复杂的命令行语法。MCP 工具的核心价值在于"一次配置，永久增强"——配置完成后，每次与 AI 对话时都可以无缝调用这些工具。

在技术实现上，MCP 工具通过标准的 JSON-RPC 协议与 AI 客户端通信，工具的功能以"工具列表"的形式暴露给 AI 模型，AI 可以按需调用。边缘核心提供了结构化的工具调用接口，使 AI 模型能够精确地理解和使用每个功能点，显著降低 AI 在工具使用上的错误率。

与传统的 API 集成相比，MCP 工具的优势在于无需编写代码——用户只需在配置文件中添加几行 JSON，即可让 AI 获得全新能力。AI Skill Hub 将边缘核心评为 AI 评分 8.0 分，属于同类工具中的优质选择。

📋 工具概览

边缘核心是一款遵循 MCP（Model Context Protocol）标准协议的 AI 工具扩展。通过 MCP 协议，它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务，实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用，都可以通过自然语言在 AI 对话中直接触发，极大提升生产效率。

GitHub Stars

⭐ 21

开发语言

Elixir

支持平台

Windows / macOS / Linux

维护状态

轻量级项目，按需更新

开源协议

NOASSERTION

AI 综合评分

8.0 分

工具类型

MCP工具

Forks

📖 中文文档

以下内容由 AI Skill Hub 根据项目信息自动整理，如需查看完整原始文档请访问底部「原始来源」。

📌 核心特色

通过标准 MCP 协议与 Claude、Cursor 等主流 AI 客户端深度集成
提供结构化工具调用接口，显著降低 AI 集成复杂度
支持 Claude Desktop 和 Claude Code 无缝接入，开箱即用
可与其他 MCP 工具组合叠加，构建完整 AI 工作站
轻量无侵入设计，不影响现有系统架构

🎯 主要使用场景

在 Claude Desktop 对话中直接调用本地工具，实现 AI 与系统的深度联动
通过自然语言驱动复杂的多步骤自动化任务，代替繁琐手动操作
将多个 MCP 工具组合使用，构建个人专属 AI 工作站

以下安装命令基于项目开发语言和类型自动生成，实际以官方 README 为准。

安装命令

# 方式一：通过 Claude Code CLI 一键安装
claude skill install https://github.com/wenet-ec/edge-core

# 方式二：手动配置 claude_desktop_config.json
{
  "mcpServers": {
    "----": {
      "command": "npx",
      "args": ["-y", "edge-core"]
    }
  }
}

# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json

📋 安装步骤说明

确认已安装 Node.js（v18 或以上版本）
打开 Claude Desktop 或 Claude Code 的 MCP 配置文件
按「交给 Agent 安装 → Claude Desktop」标签中的 JSON 配置填入 mcpServers 字段
保存配置文件并重启 Claude 客户端
重启后，在对话中即可使用本工具

以下用法示例由 AI Skill Hub 整理，涵盖最常见的使用场景。

常用命令 / 代码示例

# 安装后在 Claude 对话中直接使用
# 示例：
用户: 请帮我用 边缘核心 执行以下任务...
Claude: [自动调用 边缘核心 MCP 工具处理请求]

# 查看可用工具列表
# 在 Claude 中输入："列出所有可用的 MCP 工具"

以下配置示例基于典型使用场景生成，具体参数请参照官方文档调整。

配置示例

// claude_desktop_config.json 配置示例
{
  "mcpServers": {
    "____": {
      "command": "npx",
      "args": ["-y", "edge-core"],
      "env": {
        // "API_KEY": "your-api-key-here"
      }
    }
  }
}

// 保存后重启 Claude Desktop 生效

📑 README 深度解析真实文档完整度 33/100 查看 GitHub 原文 →

以下内容由系统直接从 GitHub README 解析整理，保留代码块、表格与列表结构。

Edge Core

Open-source self-hostable control plane for distributed Linux and Edge fleets — WireGuard mesh, SSH proxy, remote execution, Prometheus metrics, all over one REST API/MCP.

📖 Docs: wenet-ec.github.io/edge-core

Edge Core is an infrastructure management platform for fleets of Linux machines you don't physically touch. Cloud VMs across providers, on-premises servers, factory-floor equipment, Raspberry Pis, homelab boxes, IoT devices — anywhere you have N machines and want a single HTTP API to operate them. You get a secure WireGuard mesh, remote command execution, SSH without exposing port 22, HTTP/SOCKS5 forward proxying through any node, and Prometheus metrics aggregation.

We named the project "Edge Core" because the founding pain came from edge devices, but **"edge" here means any machine you don't physically touch right now** — a cloud VM in Frankfurt, a bare-metal box in a colo, or a Raspberry Pi in a factory. The control plane doesn't care; it's all the same problem.

Runs on standard Linux hosts (glibc + systemd, kernel ≥ 5.6 for built-in WireGuard). Tested on Ubuntu 22.04 / 24.04 and Debian 12 (x86_64 + ARM64); other glibc/systemd distros should work — see Host compatibility below. Self-hosted, no vendor lock-in.

The agent that runs on your machines and the Nexmaker shared library are open-source under Apache 2.0. The admin server is source-available under the Elastic License 2.0 — you can self-host, modify, and use it commercially. The one thing we reserve is the right to offer Edge Admin as a hosted service to the public; we hope you respect that decision so we can keep the rest of Edge Core fully free, with no future feature gates or surprise relicensing. See License below for details.

Getting started

Everything runs in Docker Compose — no Elixir or Go required on the host.

Pick the setup that fits your needs and follow its README:

Setup	Description	Start here
Lite	Single admin, Mosquitto broker, no metrics stack. Good for homelab, small fleets, or first deployments.	[`examples/lite/`](https://github.com/wenet-ec/edge-core/blob/main/examples/lite/README.md)
Standard	4 admin instances across 2 clusters, EMQX, full Prometheus metrics. Production-ready HA setup.	[`examples/standard/`](https://github.com/wenet-ec/edge-core/blob/main/examples/standard/README.md)

Each README covers: server requirements, configuration, enrolling your first node, and upgrading. Once your admin is running, the user guide walks through the day-to-day surface (Swagger, MCP, proxy, metrics, events).

Admin deployment matrix

The admin is always a containerized deployment target. We do not support running it directly as a bare Linux host process.

Deployment surface	Status	Notes
Docker Compose on Linux	Supported	Canonical deployment path today
Docker on Linux	Supported	Same containerized runtime model
Kubernetes	Planned / not yet shipped	`examples/k8s` is still TODO
Bare Linux host process	Unsupported	Not a supported runtime shape

The admin is much less host-sensitive than the agent: it runs containerized and uses wireguard-go (userspace) inside its own container, so it does not depend on host WireGuard kernel support.

Components

Directory	Description
`edge_admin/`	Phoenix admin server — REST API, OpenAPI, AsyncAPI, MCP server, HTTP/SOCKS5 proxies (PostgreSQL, Oban)
`edge_agent/`	Phoenix agent — embedded SSH server, HTTP/SOCKS5 proxies, metrics exporters (SQLite, Oban)
`nexmaker/`	Shared Elixir lib — Netmaker API + netclient CLI wrapper
`examples/lite/`	Single admin, Mosquitto, no metrics — good for small fleets or resource-constrained servers
`examples/standard/`	4 admins across 2 clusters, EMQX, Prometheus — when you need HA or more node capacity
`examples/relay/`	Self-hosted DERP relay node — optional, for agents behind strict NAT
`examples/sidecar/`	Agent as a sidecar container (bridge networking) rather than host-networked
`examples/event_brokers/`	NATS, Redpanda, Kafka, RabbitMQ, Redis, and MQTT compose files (AWS SNS is managed — provisioning notes inline)
`docs/`	Architecture docs and API specs

Compared to alternatives

	Edge Core	Balena	Ansible¹	Tailscale / Headscale	FleetDM
Self-hosted	✅	Partial	✅	Partial / ✅	✅
Works for general Linux	✅	❌ IoT-only	✅	✅	✅
Built-in VPN mesh	✅	❌ hub-and-spoke (OpenVPN)	❌	✅	❌
SSH proxy (no VPN client needed)	✅	❌	❌	✅	❌
HTTP forward proxy to edge	✅	❌	❌	❌	❌
Remote command execution	✅	✅	✅	❌	Partial
Prometheus metrics via admin	✅	Partial	❌	❌	❌
Works behind symmetric NAT	✅ DERP	✅	❌	✅ DERP	❌
No vendor lock-in	✅	❌	✅	❌ / ✅	✅

¹ Ansible is a complement, not a competitor — see below.

vs Balena: Balena is optimized for IoT and requires their cloud or their OS. Edge Core is designed to run on general-purpose Linux (see Host compatibility for tested distros), any cloud or on-prem. You own everything.

vs Tailscale / Headscale: Tailscale (and its self-hosted equivalent Headscale) is a polished, opinionated mesh networking product — VPN, identity, SSH, ACLs, MagicDNS, exit nodes. We use Tailscale's DERP relay protocol for our own NAT-traversal fallback, so this isn't really a head-to-head comparison. Edge Core sits one layer up: the network plumbing is solved (via Netmaker + DERP), and the product is the fleet-management layer on top — command execution, centralized SSH credentials, HTTP/SOCKS5 forward proxying, metrics aggregation, MCP. If your need is "give my team SSH'd access to my fleet over a VPN", Tailscale is probably what you want. If you also need to operate the fleet from a control plane, that's where Edge Core fits.

Works alongside Ansible: Edge Core is not an Ansible replacement — it's the network layer that makes Ansible work on unreachable machines. Nodes behind NAT or firewalls are normally inaccessible to Ansible. With Edge Core, you SSH through the admin proxy tunnel to reach them, then run your playbooks as normal. No changes to your Ansible setup needed — just configure ProxyCommand in your SSH config to route through the admin's SOCKS5 proxy and Ansible works as if the nodes were on your local network.

vs running EMQX/Mosquitto yourself: If you want edge-to-cloud communication and don't need a message bus, Edge Core gives you command delivery, result collection, SSH access, and full TCP tunneling (via HTTP and SOCKS5 forward proxy) without ever touching an MQTT broker. The broker in this repo (EMQX/Mosquitto) is internal Netmaker infrastructure — your application code never sees it.

🎯 aiskill88 AI 点评 A 级 2026-06-24

高质量的开源MCP工具，支持多种功能