keenetic-mcp

keenetic-mcp

An MCP server that lets an AI assistant manage your Keenetic router in plain language — list connected devices, pin static DHCP leases, rename devices, check WAN status, and reboot.

Category
访问服务器

README

keenetic-mcp

An MCP server that lets an AI assistant (Claude, or any MCP client) manage your Keenetic router in plain language — list connected devices, pin static DHCP leases, rename devices, check WAN status, and reboot.

It talks to the router's built-in RCI JSON API over HTTP, so there's no cloud, no third-party service, and nothing leaves your network. You run it yourself against your own router.

Tested against a Keenetic Hopper DSL (KN-3610). The RCI API is shared across the Keenetic line (Giga, Viva, Hopper, Ultra, …), so other models should work too. If you confirm one, please open an issue/PR to grow the list.

Why

Keenetic's RCI API is powerful but poorly documented, and its auth is a fiddly two-step MD5-challenge + SHA256 dance that trips people up for hours. This wraps the common device/lease/WAN operations behind clean MCP tools so you can just say "pin the Raspberry Pi to a static IP" and it happens.

Tools

Tool Type What it does
list_devices() read Active DHCP lease table (ip / mac / name / remaining time)
list_static_leases() read Static reservations (ip dhcp host in running-config)
find_free_ip(start=100, end=149) read Suggest a free IP for a new static reservation
wan_status() read WAN/internet status, WAN IP, uptime, CPU/memory
list_port_forwards() read Port-forward / static NAT rules (ip static)
set_static_lease(mac, ip, name="") write Assign a fixed IP to a MAC (low-level — you supply the MAC)
pin_device(identifier, ip="") write ⭐ Find a device by IP/MAC/name and pin its current (or a given) IP in one step. Conflict-guarded.
rename_device(identifier, name) write ⭐ Set a device's persistent display name (the name shown in the web UI)
remove_static_lease(mac) write Remove a reservation
reboot(confirm=True) write Reboot the router (only with confirm=True; internet drops ~1-2 min)

Safety: write tools are off by default

Read tools always work. Write tools are disabled until you set KEENETIC_ENABLE_WRITES=1 in the server environment. This means a fresh install can look but not touch — no accidental reboots or lease changes while you explore. Even once enabled, write tools are guarded:

  • pin_device refuses if the target IP is already active on, or statically reserved to, a different device.
  • Address-like identifiers ("66") never fuzzy-match a device name (e.g. Room-66-Cam).
  • Every write reads the RCI response back and reports status:error instead of silently "succeeding".
  • reboot requires an explicit confirm=True.

Install

Requires Python 3.9+.

git clone https://github.com/<you>/keenetic-mcp.git
cd keenetic-mcp
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt

cp .env.example .env
# edit .env: set KEENETIC_PASS (and KEENETIC_ENABLE_WRITES=1 if you want write tools)

Run it:

.venv/bin/python server.py

It starts a streamable-HTTP MCP server on http://0.0.0.0:8905/mcp (configurable via KEENETIC_HOST / KEENETIC_PORT).

Configuration

All configuration is via environment variables (see .env.example):

Variable Default Notes
KEENETIC_URL http://192.168.1.1 Your router's web-admin address
KEENETIC_USER admin Admin username
KEENETIC_PASS Required. Admin password
KEENETIC_ENABLE_WRITES 0 Set to 1 to enable write tools
KEENETIC_HOST 0.0.0.0 Bind address (use 127.0.0.1 for local-only)
KEENETIC_PORT 8905 Bind port

Connecting an MCP client

Claude Code

Add to your ~/.mcp.json (or project .mcp.json):

{
  "mcpServers": {
    "keenetic": {
      "type": "http",
      "url": "http://127.0.0.1:8905/mcp"
    }
  }
}

Then restart Claude Code (MCP config isn't hot-reloaded). Ask: "list my router's devices".

Claude Desktop / other clients

Point the client at the same streamable-HTTP URL. If your client only supports stdio, run the server with an HTTP-to-stdio bridge, or open an issue — a stdio transport option is easy to add.

Running as a service (Linux)

A keenetic-mcp.service systemd unit is included. Put the repo at /opt/keenetic-mcp, create the venv there, then:

sudo cp keenetic-mcp.service /etc/systemd/system/
sudo systemctl daemon-reload
sudo systemctl enable --now keenetic-mcp

How the auth works

For anyone reusing the RCI API directly, this is the part that's hard to find. Keenetic uses a two-step challenge:

  1. GET /auth → returns 401 with X-NDM-Realm and X-NDM-Challenge headers.
  2. Compute md5 = MD5("<user>:<realm>:<pass>"), then sha = SHA256(challenge + md5).
  3. POST /auth with {"login": user, "password": sha} → sets a session cookie.
  4. POST /rci/ with a list[dict] command tree, using that cookie.

Write commands must be followed by [{"system":{"configuration":{"save":{}}}}] or they're lost on reboot. See server.py for the full, working implementation.

Security notes

  • The server has full admin control of your router. Bind it to 127.0.0.1 (or a trusted LAN only) — never expose port 8905 to the internet.
  • Your password lives only in .env (gitignored). Nothing is sent anywhere except your own router.
  • This is not affiliated with or endorsed by Keenetic. Use at your own risk.

License

MIT — see LICENSE.

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