blog.matterxiaomi.com

Part4 - Setting up a new Zigbee network using Avatto GW70-Ti Zigbee 3.0 USB Dongle( zigbee2mqtt)

2026-10-19T20:44:54Z

Part4 - Setting up a new Zigbee network using Avatto GW70-Ti Zigbee 3.0 USB Dongle( zigbee2mqtt)

Avatto GW70-Ti Zigbee 3.0 USB Dongle

The Avatto GW70-Ti Zigbee 3.0 USB Dongle is built on the Texas Instruments CC2652P chip paired with a CP2102N USB-to-UART bridge, it is natively supported by Zigbee2MQTT using the standard zstack adapter driver.

Architecture Overview

Architecture Overview

Zigbee Sensors (Temp, Motion, Switch)->USB Zigbee Dongle (Pi)->Zigbee2MQTT (Container)->osquitto MQTT Broker (Container)->Home Assistant (Container)

step 1.Insert your USB Zigbee dongle into one of the blue USB 3.0 ports of your Raspberry Pi.

ls -l /dev/ttyUSB0
ls -la /dev/serial/by-id/

step 2.Install Zigbee2MQTT (Container)

step 3.Configure Zigbee2MQTT

web ui

http://ip:8080

step 4. add zigbee device - In the Zigbee2MQTT Web UI

Pair Your First Device

web ui

http://ip:8080

4.1 Enable Pairing Mode

navigate to http://<your-pi-ip>:8080 to access the Zigbee2MQTT Dashboard.

Permit join (All)

4.2 reset zigbee device

Xiaomi Mi door and window sensor（MCCGQ01LM）

Press and hold the reset button on the device for +- 5 seconds (until the blue light starts blinking).

https://www.zigbee2mqtt.io/devices/MCCGQ01LM.html

4.3 Auo Discovery Device by Zigbee2MQTT

useful links

AVATTO Dongle Flasher

https://dongle.avattosmart.com/

Connecting Zigbee Devices to Home Assistant Container via Docker

https://ohyaan.github.io/tips/smart_home__connecting_zigbee_devices_to_home_assistant_container_via_docker/

Part8 - Home Assistant + llama.cpp + Home-3B-v3 : Fully Local LLM Voice

2026-09-16T19:21:02Z

Raspberry Pi 5 (8GB) using Home Assistant, llama.cpp (llama-server), and a compact 3B model like Home-3B-v3 (or similar small-parameter quantized models) provides a private, offline smart home experience.

Step 1: Download a quantized model
Step 2: Run llama.cpp (llama-server) Locally with docker
Step 3: Connect Llama.cpp server to Home Assistant

Step 1: Download a quantized model

# 1. https://gub.com/acon96/home-llmith/Home-Llama-3.2-3B
# Home Llama 3.2 3B (Latest Stable - Best Quality & Multilingual)
wget -O Home-Llama-3.2-3B-Q4_K_M.gguf \
  "https://huggingface.co/acon96/Home-Llama-3.2-3B/resolve/main/Home-Llama-3.2-3B.q4_k_m.gguf"

# 2. https://gub.com/acon96/home-llmith/acon96/Home-FunctionGemma-270m
# Home FunctionGemma 270M (Ultra-fast / Low RAM - ideal for sub-second responses)
wget -O Home-FunctionGemma-270m-Q4_0.gguf \
  "https://huggingface.co/acon96/Home-FunctionGemma-270m/resolve/main/Home-FunctionGemma-270m.q4_0.gguf"

# 3. Home 3B v3 (StableLM-Zephyr base)
wget -O Home-3B-v3-Q4_K_M.gguf \
  "https://huggingface.co/acon96/Home-3B-v3-GGUF/resolve/main/Home-3B-v3.q4_k_m.gguf"

Step 2: Run llama.cpp (llama-server) Locally with docker

docker run -it --rm \
  --name homellm \
  -v /datadocker/home-llm/models:/models \
  -p 8092:8080 \
  ghcr.io/ggml-org/llama.cpp:server-b10499  \
  -m /models/Home-3B-v3-Q4_K_M.gguf \
  --host 0.0.0.0 \
  --port 8080 \
  --threads 4 \
  --jinja

vist test:

http://192.168.2.125:8092/

Step 3: Connect Llama.cpp server to Home Assistant

Connect Home Assistant to llama.cpp server using a llama.cpp integration.

add llama.cpp integration

URL:http://192.168.2.125:8092/v1

part 1 - Claude Code + ha mcp server for Home Assistant

2026-08-29T18:31:44Z

I use Claude Code (Anthropic’s AI assistant) directly into Home Assistant. With it, you can create automations, debug configurations, and manage your smart home using natural language through Claude Code.

What does it do?
Usage examples
How does it work?
Requirements

What does it do?

Create automations by describing what you want in plain English

Access your entities - Claude can see your sensors, lights, climate, etc.

Edit YAML configurations directly

Debug automations that aren’t working

Usage examples

claude “List all my automations”

claude “Create an automation to turn off lights when I leave home”

claude “Why isn’t my motion sensor automation working?”

claude “Show the state of all temperature sensors”

How does it work?

The HA-MCP Custom Component from HACS uses ha-mcp server(Model Context Protocol) which allows Claude to:

Query entity states

Control devices

List available services

Access the entire Home Assistant API

Requirements

Home Assistant Docker

Anthropic account

part2 - Configure the Home Assistant MCP Server for your AI assistant

2026-08-29T22:25:26Z

Combining Claude Code with an MCP Server is the modern, safest, and most powerful way to manage Home Assistant using natural language instead of relying strictly on manual SSH commands.

The best way to start is by using the HA-MCP Custom Component from HACS.

step 1. Install HA-MCP Custom Component from HACS, then restart Home Assistant.

step 2.Start HA-MCP Server in Home Assistant.

After it's running: you can get the HA-MCP sidebar panel.

sterp 3.Copy the connect URL from the HA-MCP Custom Componen's Configure screen

step 4.Paste URL into your AI client

Connecting Claude Code to Home Assistant with HA-MCP Server

step 5.write code

Claude Code is a general AI agent that happens to write code.

Haiku is fine for home assistant,

useful links

Claude Code

https://homeassistant-ai.github.io/ha-mcp/setup/?method=ha-component&client=claude-code

Connect Claude Code to tools via MCP

https://code.claude.com/docs/en/mcp

part 7 - Build Your Own Private AI Station: Llama server with Docker+ Open WebUI with Docker

2026-08-20T15:18:45Z

I am running llama.cpp server with Gemma 4 (E2B) model on an rp55 8GB.

Connecting llama.cpp running Gemma 4 (E2B) to Home Assistant via HA-MCP allows the AI model to perform local, tool-calling smart home control.

RPi5 8GB +docker llama.cpp server + Gemma 4 E2B + docker Home Assistant + ha-mcp-integration(build-in ha-mcp server )+docker Open WebUI.

Open WebUI will begin using your local Llama.cpp server as a backend!

Overview
How it work?
Install
Connections
- Step 4: Connect Llama.cpp server via Open WebUI
- Step 5: Connect ha-mcp server via Open WebUI

Overview

Runs the full ha-mcp server inside Home Assistant.Transport: use Streamable HTTP

Set up your Llama.cpp server with docker

Llama.cpp server Load large models locally

Enable 4GB of ZRAM swap space to prevent hard OOM crashes during heavy context processing.

Integrate with Open WebUI for a seamless interface

#docker images
REPOSITORY                      TAG       IMAGE ID       CREATED         SIZE
ghcr.io/ggml-org/llama.cpp      server    bde659bfc300   8 days ago      1.21GB
ghcr.io/open-webui/open-webui   v0.11.0   72c0ba641ba7   2 weeks ago     7.15GB

How it work?

Architecture Overview

Open WebUI Docker(port 3000)
    ↓ OpenAI-compatible API
llama.cpp server (port 8080) ← Gemma 4 E2B Q4
    ↓ (tool calls via MCP)
ha-mcp (inside HA or add-on)
    ↓
Home Assistant Docker

┌─────────────────┐       MCP       ┌─────────────────┐      REST/WS      ┌─────────────────┐
│ MCP Host Client │ ──────────────> │  HA-MCP Server  │ ────────────────> │ Home Assistant  │
│ (Open WebUI /   │                 │   (:9584)       │                   │   (:8123)       │
│  MCP Agent)     │                 └─────────────────┘                   └─────────────────┘
└────────┬────────┘
         │ OpenAI API
         ▼
┌─────────────────┐
│  llama.cpp      │ (Gemma 4 E2B)
│  (:8080)        │
└─────────────────┘

Note

1.llama.cpp acts as an inference server (provider)

2.HA-MCP acts as an MCP tool provider

3.Open WebUI sends user prompts to llama.cpp server

Install

Step 1: Install Llama.cpp server with docker

Step 2: Download a Supported Model

cd /datadocker/llama-cpp

docker pull ghcr.io/ggml-org/llama.cpp:server-b10499

docker run \
  --name llama \
  --restart=unless-stopped \
  -v /datadocker/llama-cpp/models:/models \
  -p 8091:8080 \
  ghcr.io/ggml-org/llama.cpp:server:server-b10499 \
  -m /models/google_gemma-4-E2B-it-Q4_0.gguf \
  --host 0.0.0.0 \
  --port 8080 \
  --threads 4 \
  --jinja \
  --reasoning-budget 0 \
  --chat-template-kwargs '{"enable_thinking": false}' \
  --ui-mcp-proxy

you will get api:

http://192.168.2.134:8091/v1

Step 2: Install Open WebUI with docker

docker pull ghcr.io/open-webui/open-webui:v0.11.0

cd /datadocker/open-webui/

docker run -d -p 3000:8080 --name open-webui \
 -v /datadocker/open-webui/data:/app/backend/data \
 ghcr.io/open-webui/open-webui:v0.11.0


web ui

http://192.168.2.125:3000/

Step 3: Install ha-mcp-integration(build-in ha-mcp server ) in home assistant

first,Install HA-MCP Custom Component integration in ha

then Add HA-MCP Server in HA-MCP Custom Component integration.

you will get api:

http://192.168.2.125:9584/private_QQz2CLU3n8xR

Connections

Step 4: Connect Llama.cpp server via Open WebUI

connect to a AI Provider

Add the API Connection in Open WebUI

Connect Llama.cpp server to Open WebUI

Open WebUI will begin using your local Llama.cpp server as a backend

Open Open WebUI in your browser

Go to ⚙️ Admin Settings → Connections → OpenAI.

Click ➕ Add Connection.

http://192.168.2.134:8091/v1

add http://192.168.2.134:8091/v1,it is llama.cpp with gemma4 model

Step 5: Connect ha-mcp server via Open WebUI

add ha mcp server

Open Open WebUI in your browser

Go to ⚙️ Admin Settings → Tools → Integrations.

Click ➕ Add External Tool Servers.

http://192.168.2.125:9584/private_QQz2CLU3n8xR

Note

you can get Connect URL via HA-MCP Custom Component integration in ha.

http://192.168.2.125:8123/config/integrations/integration/ha_mcp_tools

debug step by step

http://192.168.2.134:8091/v1

cqrs

0.00.013.586 W srv  llama_server: CORS is set to allow all origins ('*') and no API key is set
0.00.013.586 W srv  llama_server: this can be a security risk (cross-origin attacks)
0.00.013.587 W srv  llama_server: more info: https://github.com/ggml-org/llama.cpp/pull/25655

useful links

https://docs.openwebui.com/getting-started/quick-start/connect-a-provider/starting-with-llama-cpp/

add ha mcp server

https://homeassistant-ai.github.io/ha-mcp/setup/?method=ha-component&client=open-webui&scope=local

Docker HA + HA Custom Component (in-process server)+ Docker Open WebUI + Local access on same host.

https://homeassistant-ai.github.io/ha-mcp/setup/?method=ha-component&client=open-webui&scope=local

Part 12 - update Raspberry Pi OS Trixie (2025) and increase swap to 4 GB

2026-08-15T19:41:27Z

Raspberry Pi OS Trixie (2025)

Because the RPi 5 shares system RAM between the CPU and AI workloads, careful configuration is required to prevent out-of-memory (OOM) crashes.

System Requirements
Raspberry Pi OS Trixie (2025) uses rpi-swap:
Check Current Swap Configuration

Expand Swap Using dphys-swapfile (Bookworm and Older).

Expand Swap Using rpi-swap (Trixie and Newer).

Raspberry Pi OS Trixie (2025) or newer, the system uses rpi-swap instead of dphys-swapfile.

rpi-swap is a new package that replaces/deprecates dphys-swapfile for configuring swap on your Raspberry Pi.it is a standard swap management solution, providing better performance and reliability than the old dphys-swapfile approach.

Supports: Compressed RAM swap (zram), file-based swap, or hybrid modes.

Using zram avoids small/frequent writes to /var/swap so as to increase SD card longevity.

Raspberry OS Version

cat /etc/os-release
PRETTY_NAME="Debian GNU/Linux 13 (trixie)"
NAME="Debian GNU/Linux"
VERSION_ID="13"
VERSION="13 (trixie)"
VERSION_CODENAME=trixie
DEBIAN_VERSION_FULL=13.4
ID=debian
HOME_URL="https://www.debian.org/"
SUPPORT_URL="https://www.debian.org/support"
BUG_REPORT_URL="https://bugs.debian.org/"

System Requirements

Raspberry Pi OS Trixie (2025) or newer

rpi-swap is installed.

Raspberry Pi OS Trixie (2025) uses rpi-swap:

Config Directory: /etc/rpi/swap.conf (drop-in configuration files).

Default Swap Type: Compressed RAM swap (zram).

Default Swap File: /dev/zram0.

Default Size: Up to 2048MB, sized to match RAM (capped at 50% of disk space).

Supports: Compressed RAM swap (zram), file-based swap, or hybrid modes.

rpi-swap

rpi-swap GitHub Repository (Official):https://github.com/raspberrypi/rpi-swap

Modern swap management for Raspberry Pi systems, designed to replace dphys-swapfile with a more flexible and efficient approach.

Check Current Swap Configuration

Check total swap size and usage:

free -h

 free -h
               total        used        free      shared  buff/cache   available
Mem:           7.9Gi       5.7Gi       342Mi       2.7Mi       2.0Gi       2.2Gi
Swap:          2.0Gi       2.0Gi       656Ki

Check swap details

cat /proc/meminfo | grep Swap

 cat /proc/meminfo | grep Swap
SwapCached:           48 kB
SwapTotal:       2097136 kB
SwapFree:          26944 kB

see where the swap file is located:

swapon --show

 swapon --show
NAME       TYPE      SIZE USED PRIO
/dev/zram0 partition   2G   0B  100

rpi-swap uses systemd generators so 'rpi-zram-writeback' will not appear in the list of package files.

tree /run/systemd/generator/

 tree /run/systemd/generator/
/run/systemd/generator/
├── boot-firmware.mount
├── dev-zram0.swap
├── getty.target.wants
│   └── serial-getty@ttyAMA10.service -> /usr/lib/systemd/system/serial-getty@.service
├── local-fs.target.requires
│   ├── boot-firmware.mount -> ../boot-firmware.mount
│   └── -.mount -> ../-.mount
├── local-fs.target.wants
│   ├── systemd-fsck-root.service -> /usr/lib/systemd/system/systemd-fsck-root.service
│   └── systemd-remount-fs.service -> /usr/lib/systemd/system/systemd-remount-fs.service
├── -.mount
├── netplan.stamp
├── rpc_pipefs.target
├── rpi-setup-loop@var-swap.service.d
│   ├── fast-symlink.conf
│   ├── no-dependency-cycle.conf
│   └── resize-swap.conf
├── rpi-zram-writeback.timer
├── run-rpc_pipefs.mount
├── sockets.target.wants
│   └── sshd-unix-local.socket -> ../sshd-unix-local.socket
├── sshd-unix-local@.service -> /usr/lib/systemd/system/sshd@.service
├── sshd-unix-local.socket
├── swap.target.wants
│   └── dev-zram0.swap -> ../dev-zram0.swap

install/configure rpi-swap (Trixie only):

sudo apt install rpi-swap
sudo reboot

Configuration files are located at:

/etc/rpi/swap.conf

detailed configuration options

man swap.conf

verify that swap has been configured properly:

systemctl list-units --type swap

systemctl list-units --type swap
  UNIT           LOAD   ACTIVE SUB    DESCRIPTION                             
  dev-zram0.swap loaded active active rpi-swap managed swap device (zram+file)

Legend: LOAD   → Reflects whether the unit definition was properly loaded.
        ACTIVE → The high-level unit activation state, i.e. generalization of SUB.
        SUB    → The low-level unit activation state, values depend on unit type.

1 loaded units listed. Pass --all to see loaded but inactive units, too.
To show all installed unit files use 'systemctl list-unit-files'.

References

rpi-swap GitHub Repository (Official):https://github.com/raspberrypi/rpi-swap

run-local-llm-server-part6 debian vmware on windows 10

2026-08-23T11:26:51Z

Llama.cpp is a high-performance inference engine written in C/C++ designed for running Llama and compatible models in the GGUF format.

vmware on windows 10

Overview

Set up your Llama.cpp server

Load large models locally

docker pull ghcr.io/ggml-org/llama.cpp:server-b10499
docker pull ghcr.io/open-webui/open-webui:v0.11.0

Step 1: Download a model in GGUF format from Hugging Face.

# llama.cpp only works with GGUF

# Create a dedicated folder for models:
mkdir -p /datadocker/llama-cpp

cd /datadocker/llama-cpp

mkdir models

cd /datadocker/llama-cpp/models


https://huggingface.co/unsloth/gemma-4-E2B-it-GGUF/tree/main

https://huggingface.co/bartowski/google_gemma-4-E2B-it-GGUF/tree/main

google_gemma-4-E2B-it-Q4_0.gguf

wget https://huggingface.co/bartowski/google_gemma-4-E2B-it-GGUF/resolve/main/google_gemma-4-E2B-it-Q4_0.gguf?download=true

Step 2: Install Llama.cpp server with docker

cd /datadocker/llama-cpp

docker run \
  --name llama \
  --restart=unless-stopped \
  -v /datadocker/llama-cpp/models:/models \
  -p 8091:8080 \
  ghcr.io/ggml-org/llama.cpp:server \
  -m /models/google_gemma-4-E2B-it-Q4_0.gguf \
  --host 0.0.0.0 \
  --port 8080 \
  --threads 4 \
  --jinja \
  --reasoning-budget 0 \
  --chat-template-kwargs '{"enable_thinking": false}' \
  --webui-mcp-proxy

Step 4: Connect Llama.cpp server via WebUI

part4 - Claude Code for Home Assistant(ha-mcp Server directly inside Home Assistant)

2026-08-29T18:32:34Z

Claude Code for Home Assistant

Is there a guide to setup home assistant with MCP and integrate with Claude？

I created an guide that integrates Claude Code (Anthropic’s AI assistant) directly into Home Assistant. With it, you can create automations, debug configurations, and manage your smart home using natural language through a browser-based terminal.

Install type
Usage examples
How does it work?
Requirements
Inside home assistant
- Installation ha-mcp-integration
Copy the connect URL for AI Client
Verify the Target Server is Active:

Install type

Run MCP Server directly inside Home Assistant.

Note

Newly solutions usually ran as an internal Python process inside Home Assistant.

ha-mcp-integration is a native Home Assistant integration that embeds an MCP (Model Context Protocol) Server directly inside Home Assistant.

offcial website:https://github.com/homeassistant-ai/ha-mcp-integration

download:https://codeload.github.com/homeassistant-ai/ha-mcp-integration/zip/refs/tags/v1.3.0

Usage examples

claude “Create an automation to turn off lights when I leave home”

Create an automation that turns on the bedroom AC when the temperature exceeds 28°C.

How does it work?

The ha-mcp-integration uses ha-mcp (Model Context Protocol) server which allows Claude to:

Query entity states

Control devices

List available services

Access the entire Home Assistant API

Besides the built-in tools, the server can expose tools coming from MCP servers.

ha-mcp-integration

The integration starts an MCP Server inside Home Assistant..

The integration registers MCP Tools

AI Mcp client connects via MCP.

Architecture

                 AI Client
       (ChatGPT / Claude / Cursor)
                    │
          MCP Protocol (HTTP/SSE)
                    │
        ┌────────────────────────┐
        │ Home Assistant         │
        │ ha-mcp-integration     │
        │                        │
        │  MCP Server            │
        │  Tool Registry         │
        │  Authentication        │
        └──────────┬─────────────┘
                   │
          Home Assistant Core
                   │
      ┌────────────┴─────────────┐
      │ States                   │
      │ Services                 │
      │ Devices                  │
      │ Areas                    │
      │ Automations              │
      │ Scripts                  │
      │ History                  │
      └──────────────────────────┘

ha-mcp-integration:Full MCP Server running inside Home Assistant. AI clients connect directly to it.

ha_mcp_tools:A library of MCP tools used by the MCP Server.

ha-mcp-integration
        │
        ├── starts MCP Server
        │
        └── uses ha_mcp_tools
                 │
                 ├── call_service
                 ├── get_state
                 ├── list_entities
                 ├── search_devices
                 └── ...

ha-mcp-integration does not "think up" automations by itself. Instead, it exposes Home Assistant capabilities as MCP tools, allowing an AI (ChatGPT, Claude, Cursor, etc.) to inspect your Home Assistant configuration and then create or modify automations.

Requirements

Home Assistant Docker

Anthropic account

Inside home assistant

Installation ha-mcp-integration

ha-mcp-integration=HA-MCP Custom Component

Step 1: Download the HA-MCP Custom Component:

https://github.com/homeassistant-ai/ha-mcp-integration/releases

wget https://github.com/homeassistant-ai/ha-mcp-integration/archive/refs/tags/v1.3.0.zip

copy custom_components/ha_mcp_tools/ into config/custom_components/

Step 2: Restart home assistant:

Step 3: Install the HA-MCP Custom Component:

Choose Add HA-MCP Server.it

Copy the connect URL for AI Client

http://192.168.2.125:8123/config/integrations/integration/ha_mcp_tools

Copy the connect URL from the entry's Configure screen (Settings → Devices & Services → HA-MCP Custom Component → HA-MCP Server → Configure)

Verify the Target Server is Active:

http://192.168.2.125:9584/private_QQz2CLU3n8xRQ1

HA-MCP server is up and running!

To connect, paste the full URL (including the /private_... key) into the
connector or MCP settings of your AI/LLM client. No username or password required.
Setup instructions: https://homeassistant-ai.github.io/ha-mcp/

--- Seeing this page? Your URL is set up correctly ---

If this page loads in your browser, the MCP server is reachable and the
URL is correct. If your AI client still cannot connect, the problem is
not on HA-MCP's side. Common causes:

- Geo / country blocking in your reverse proxy / CDN (Cloudflare, NGINX,
  Traefik, Zoraxy, etc.). Most AI/LLM services connect from US-based
  cloud infrastructure, so if you block US IP addresses (or only allow
  your own country), that is why your client cannot connect. Allow your
  provider's IP ranges (or your client's egress IPs). For example,
  Claude.ai connects from Anthropic's network, 160.79.104.0/21.
- WAF, bot-blocking, or rate-limiting rules on the proxy that drop or
  challenge the request.
- The AI client's network can sometimes be spotty -- you may just need
  to try connecting again.
- The AI client itself refusing certain domains or proxy providers on
  its end. This is rare and outside your control; try a different
  hostname or proxy if you suspect it.

Your proxy's access logs will show the blocked attempt -- look for the
request from your AI provider's IP (e.g. an Anthropic 160.79.x.x address).

--- Cloudflare Users ---

If your LLM cannot connect, Cloudflare's "Block AI training bots"
setting is the most common cause. To disable it:

1. Log in to Cloudflare (https://dash.cloudflare.com)
2. In the left sidebar, click Domains, then click Overview
3. Click on the domain you use for connecting to Home Assistant
4. On the right side, find "Control AI Crawlers"
5. Under "Block AI training bots", open the dropdown
6. Select "do not block (allow crawlers)"

Screenshot of the setting:
https://homeassistant-ai.github.io/ha-mcp/images/cloudflare-ai-crawlers-setting.jpg

part3 - Using Claude Code to manage your Home Assistant config step by step(docker ha-mcp server))

2026-08-29T17:29:38Z

Install type

Run Standalone ha-mcp MCP Server outside Home Assistant.

Note

Earlier solutions (such as the standalone ha-mcp project) usually ran as an external Python process:

Claude
      │
Standalone ha-mcp Server
      │
REST/WebSocket
      │
Home Assistant

This project provides a seamless integration between Home Assistant and Claude AI via the Model Context Protocol (MCP), enabling using Claude Code to manage your Home Assistant config through natural language step by step.

Install type
Prerequisites
Architecture
Installation
Connecting Claude Code to Home Assistant's HA-MCP server
- step 3.Create a long-lived access token in Home Assistant
- step 4.Configure Claude
Usage Examples

Prerequisites

Home Assistant ha-mcp MCP Server

https://github.com/homeassistant-ai/ha-mcp

Integration

https://github.com/homeassistant-ai/ha-mcp-integration

AI client

Claude Desktop

Architecture

┌─────────────────┐
│   Claude AI     │
│  (Sonnet 4.5)   │
└────────┬────────┘
         │ MCP Protocol
         │
┌────────▼────────┐
│ Standalone  MCP Server    │
│  (WebSocket)    │
└────────┬────────┘
         │ REST API
         │
┌────────▼────────┐
│ Home Assistant  │
│   (v2026.7)    │
└─────────────────┘

MCP server connects to my HA instance through a Long-lived access token to access the Home Assistant API, giving it the ability to search entities, read automation traces, reload configs, and query device states.

ha-mcp is a standalone server built for configuring, building, and debugging your smart home

Installation

step 1.Start MCP server

Runs the ha-mcp server.

docker pull ghcr.io/homeassistant-ai/ha-mcp:latest


docker run -d \
  --name ha-mcp \
  --restart unless-stopped \
  -p 9000:9000 \
  -e HOMEASSISTANT_URL=http://192.168.1.100:8123 \
  -e HOMEASSISTANT_TOKEN=xxxxxxxxxxxxxxxx \
  -e MCP_PORT=9000 \
  ghcr.io/homeassistant-ai/ha-mcp:latest \
  ha-mcp-web

Docker (HTTP server): run ghcr.io/homeassistant-ai/ha-mcp in HTTP mode, pointed at your Home Assistant URL and a long-lived token.

Start Docker Container

docker run -d --name ha-mcp -p 8086:8086 -v ha-mcp-data:/home/mcpuser/.ha-mcp -e HOMEASSISTANT_URL={{HOMEASSISTANT_URL}} -e HOMEASSISTANT_TOKEN={{HOMEASSISTANT_TOKEN}} ghcr.io/homeassistant-ai/ha-mcp:latest ha-mcp-web

Server will be at: http://YOUR_IP:8086/mcp

detail:https://homeassistant-ai.github.io/ha-mcp/setup/?method=docker&client=claude-desktop&scope=local&platform=linux

step 2.Install HA-MCP Custom Component in Home Assistant

To add the Model Context Protocol Server service to your Home Assistant instance, use this Integration:

https://github.com/homeassistant-ai/ha-mcp-integration

source code:https://github.com/homeassistant-ai/ha-mcp/tree/master/custom_components/ha_mcp_tools

Connecting Claude Code to Home Assistant's HA-MCP server

step 3.Create a long-lived access token in Home Assistant

You may create a Long-lived access token to allow the client to access the Home Assistant API.

Go to User profile > Security tab,select Create token.

step 4.Configure Claude

An LLM application acts as a client and can connect to multiple MCP servers to provide context.

{
  "mcpServers": {
    "homeassistant": {
      "type": "http",
      "url": "http://localhost:9090",
      "headers": { "Authorization": "Bearer your-ha-access-token" }
    }
  }
}

Usage Examples

Automation Creation

"Create an automation that turns off all lights when I leave home"
"Set up a morning routine at 7 AM with lights and coffee machine"
"Alert me when the washing machine finishes"

useful links

https://homeassistant-ai.github.io/ha-mcp

download stable release from the latest release

https://github.com/homeassistant-ai/ha-mcp/releases

https://github.com/homeassistant-ai/ha-mcp/releases/tag/v8.0.0

Home Assistant MCP Server Add-on doc

https://github.com/homeassistant-ai/ha-mcp/blob/master/homeassistant-addon/README.md

Runs the full ha-mcp server inside Home Assistant.

Home Assistant MCP Setup：https://homeassistant-ai.github.io/ha-mcp/setup/?method=ha-component&client=open-webui&scope=local

LM Studio +ollama+unofficial HA-MCP Server+ Claude Code

https://www.xda-developers.com/i-connected-my-local-llm-to-home-assistant-through-mcp/

Ha Template

2026-07-17T18:55:21Z

The Template integrationIntegrations connect and integrate Home Assistant with your devices, services, and more. [Learn more] allows creating entities which derive their values from other data.

Legacy template entities (alarm_control_panel, binary_sensor, cover, fan, light, lock, sensor, switch, vacuum, weather) have been removed in 2026.6. You have to migrate to modern template entity syntax.

ha 2026.5
ha 2026.6

ha 2026.5

Old Syntax

#
# ---------------------------------- vacuum Template -----------------------------------
#
# https://www.home-assistant.io/integrations/vacuum.template/  old
# https://www.home-assistant.io/integrations/template/#state-based-vacuum---custom-attributes  new
vacuum:
  - platform: template
    vacuums:
      virtual_hallway_room_vacuum:
      #  unique_id:  virtual_hallway_room_vacuum

        start:
            action: script.vacuum_start
        pause:
            action: script.vacuum_pause
        stop:
            action: script.vacuum_stop
        return_to_base:
            action: script.vacuum_return_to_dock
        clean_spot:
            action: script.vacuum_clean_spot
        locate:
            action: script.vacuum_locate_vacuum
        set_fan_speed:
            action: script.vacuum_set_fan_speed
            data:
              speed: "{{ fan_speed }}"
        fan_speeds:
            - Low
            - Medium
            - High
        value_template: "{{ states('sensor.deebotx5pro') }}"
        # battery_level_template: "{{ states('sensor.deebotx5pro_battery')|int }}"
        fan_speed_template: "{{ states('sensor.vacuum_fan_speed') }}"
        attribute_templates:
          status: >-
            {% if (states('sensor.robot_vacuum_robot_cleaner_movement') == "after" and states('sensor.robot_vacuum_robot_cleaner_cleaning_mode') == "stop")  %}
              Charging to Resume
            {% elif states('sensor.vacuum.deebotx5pro') == "auto" %}
              Cleaning
            {% else %}
              Charging
            {% endif %}

Legacy (removed in 2026.6)

Unsupported YAML configuration for the command_line

Configuring the template integration by adding platform: template under the vacuum: key is not supported. The template integration must be configured under its own template: key instead.

To resolve this:

Remove the following from your YAML configuration file:
vacuum:
  - platform: template
Move the configuration under the template: key instead.

Restart Home Assistant.

ha 2026.6

New Syntax

#
# ---------------------------------- Template -----------------------------------
#
# https://www.home-assistant.io/integrations/template
# https://community.home-assistant.io/t/how-to-define-area-in-configuration-yaml/240296/13
template:
  # https://www.home-assistant.io/integrations/template/#vacuum   new template 20260711
  - vacuum:
      - name: Living Room Vacuum
        clean_segments:
          action: script.vacuum_start
          data:
            segment_ids: "{{ segment_ids }}"
        segments: >-
          {{ [
            {'id': '1', 'name': 'Kitchen'},
            {'id': '2', 'name': 'Living room', 'group': 'Upstairs'},
          ] }}
        start:
          action: script.vacuum_start
        unique_id: living_room_vacuum

useful links

https://smarthomeassistant.co.uk/articles/home-automation/home-assistant-2026-6-template-migration

blog.matterxiaomi.com

Part4 - Setting up a new Zigbee network using Avatto GW70-Ti Zigbee 3.0 USB Dongle( zigbee2mqtt)

Table of Contents

Architecture Overview

Part8 - Home Assistant + llama.cpp + Home-3B-v3 : Fully Local LLM Voice

Table of Contents

Step 1: Download a quantized model

Step 2: Run llama.cpp (llama-server) Locally with docker

Step 3: Connect Llama.cpp server to Home Assistant

part 1 - Claude Code + ha mcp server for Home Assistant

Table of Contents

What does it do?

Usage examples

How does it work?

Requirements

part2 - Configure the Home Assistant MCP Server for your AI assistant

part 7 - Build Your Own Private AI Station: Llama server with Docker+ Open WebUI with Docker

Table of Contents

Overview

How it work?

Install

Step 1: Install Llama.cpp server with docker

Step 2: Install Open WebUI with docker

Step 3: Install ha-mcp-integration(build-in ha-mcp server ) in home assistant

Connections

Step 4: Connect Llama.cpp server via Open WebUI

Step 5: Connect ha-mcp server via Open WebUI

Part 12 - update Raspberry Pi OS Trixie (2025) and increase swap to 4 GB

Table of Contents

System Requirements

Raspberry Pi OS Trixie (2025) uses rpi-swap:

Check Current Swap Configuration

run-local-llm-server-part6 debian vmware on windows 10

Overview

part4 - Claude Code for Home Assistant(ha-mcp Server directly inside Home Assistant)

Table of Contents

Install type

Usage examples

How does it work?

Requirements

Inside home assistant

Installation ha-mcp-integration

Copy the connect URL for AI Client

Verify the Target Server is Active:

part3 - Using Claude Code to manage your Home Assistant config step by step(docker ha-mcp server))

Install type

Table of Contents

Prerequisites

Architecture

Installation

step 1.Start MCP server

step 2.Install HA-MCP Custom Component in Home Assistant

Connecting Claude Code to Home Assistant's HA-MCP server

step 3.Create a long-lived access token in Home Assistant

step 4.Configure Claude

Usage Examples

Ha Template

Table of Contents

ha 2026.5

ha 2026.6