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.
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_devicerefuses 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:errorinstead of silently "succeeding". rebootrequires an explicitconfirm=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:
GET /auth→ returns401withX-NDM-RealmandX-NDM-Challengeheaders.- Compute
md5 = MD5("<user>:<realm>:<pass>"), thensha = SHA256(challenge + md5). POST /authwith{"login": user, "password": sha}→ sets a session cookie.POST /rci/with alist[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.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.