unifi-mcp

unifi-mcp

Read-only MCP server that surfaces UniFi network data from a UCG-Fiber console to Claude, enabling monitoring of hosts, sites, devices, clients, ISP metrics, and device stats via read-only API tools.

Category
Visit Server

README

unifi-mcp

A lean, read-only MCP server that surfaces UniFi network data (a UCG-Fiber console) to Claude over stdio. Written against Ubiquiti's official API key (X-API-KEY); every tool is a GET, so it can only read — never change — your network.

Tools

Tool Returns API surface
list_hosts Consoles on the account (model, IP, firmware, online state) Site Manager (cloud)
list_sites Sites with device-count statistics Site Manager (cloud)
list_devices All adopted devices: status, firmware, update state Site Manager (cloud)
get_isp_metrics(period, duration) WAN health — latency, throughput, packet loss, uptime, per WAN Site Manager (cloud)
list_clients(site_id?) Connected clients (name, IP, MAC, connection type) Connector proxy (Owner key)
get_device_stats(site_id?) Live per-device CPU, memory, load, uptime, uplink throughput Connector proxy (Owner key)

get_isp_metrics defaults to period="5m", duration="24h"; use period="1h" with duration="7d"/"30d" for longer windows. The proxy tools default to the console's local site.

Setup

Requires Python 3.11+ (the system 3.9 is too old for mcp). Uses uv.

cd unifi-mcp
uv venv --python 3.11
uv pip install mcp httpx

Configuration (environment)

  • UNIFI_API_KEY (required) — your UniFi API key. Create it at unifi.ui.com → profile → API Keys, or locally under UniFi Network → Settings → Control Plane → Integrations.
    • The proxy tools (list_clients, clients_per_device, get_device_stats) require an Owner key. The cloud tools work with any key, including a read-only Viewer key.
  • UNIFI_CONSOLE_ID (required for the proxy tools) — your console's id. Find it as the host id in the cloud GET /hosts response (or run list_hosts).
  • UNIFI_SITE_ID (required for the proxy tools) — the local site id. Find it as data[0].id in the proxy GET /sites response (often the "Default" site).

None of these are hardcoded — all are read from the environment at runtime, and the API key is never logged, committed, or written to disk by this code. The cloud-only tools (list_hosts, list_sites, list_devices, get_isp_metrics) need just UNIFI_API_KEY.

Register with Claude Code

claude mcp add unifi --scope user \
  -e UNIFI_API_KEY="$UNIFI_API_KEY" \
  -e UNIFI_CONSOLE_ID="$UNIFI_CONSOLE_ID" \
  -e UNIFI_SITE_ID="$UNIFI_SITE_ID" \
  -- /ABS/PATH/unifi-mcp/.venv/bin/python -m unifi_mcp.server

Registering against the absolute venv interpreter (not uv run) keeps cold-start to a bare exec. --scope user keeps the registration (and the key) out of any project repo.

Skill: network health

skills/network-health/ is an optional Claude Code skill that turns the tools into a one-shot health summary — WAN/ISP status, offline devices, pending firmware updates, and client distribution per AP. Install it by copying the folder into your skills dir:

cp -r skills/network-health ~/.claude/skills/

Then ask Claude "how's my network?" and it orchestrates the relevant tools.

Tests

Two layers, both runnable without touching production state (everything is a read):

# Offline: projection/envelope logic pinned to real captured payloads (no key)
.venv/bin/python test_projections.py

# Live: exercises the tools against the real API; prints raw-keys -> kept so
# field-name mismatches are visible (needs UNIFI_API_KEY; proxy ops also need
# UNIFI_CONSOLE_ID and UNIFI_SITE_ID)
UNIFI_API_KEY=... UNIFI_CONSOLE_ID=... UNIFI_SITE_ID=... .venv/bin/python smoke_test.py

Design notes

  • stdio transport — Claude launches the process on demand; zero idle cost.
  • Read-only enforced in code — the HTTP client only ever issues GET (asserted), so the read-only guarantee holds even if Ubiquiti later ships write endpoints.
  • Token-efficient — responses are projected down to monitoring-relevant fields (projections.py); ISP metrics are summarized per-WAN rather than returned raw.
  • Two API envelopes — the cloud API uses cursor pagination (pageSize/nextToken); the connector proxy uses offset pagination (offset/limit) and some endpoints (e.g. statistics/latest) return a bare object with no envelope. client.py handles all three shapes.

Recommended Servers

playwright-mcp

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.

Official
Featured
TypeScript
Magic Component Platform (MCP)

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.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

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.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

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.

Official
Featured
TypeScript
Kagi MCP Server

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.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

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.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured