MCP Assist for Home Assistant (Marketplace integration)

Installs: 294 sharing analytics
GitHub stars: 113
Connection: Works locally
Upkeep: Updated in the last 6 months
Version: v0.17.2, August 2, 2026
Category: AI and voice, AI models
Kind: Marketplace integration
Needs: Needs the Marketplace (HACS on 2026.10 and older)
Domain: mcp_assist
Source: github.com/mike-nott/mcp-assist
Issues: Report an issue

Independent directory, made by the developer of Vulpo. Not affiliated with Home Assistant or the Open Home Foundation.

About

MCP Assist is a conversation agent for Home Assistant that finds entities through MCP tools instead of sending the full entity list. It starts an MCP server inside Home Assistant, and the connected language model calls tools to look up areas, domains, entity details, and states only as a request needs them. The same tools let the model control devices, run scripts, and trigger automations. An optional smart entity index gives the model a short overview of areas, domains, people, zones, and scripts, and a setting limits how many entities each discovery returns. It works with local servers such as LM Studio, llama.cpp, and Ollama, and with cloud services such as OpenAI, Gemini, Claude, and OpenRouter. Several profiles can run side by side with different models, optional web search is available, and the interface is localized in 21 languages.

Automatic summary of the README and code of mike-nott/mcp-assist. Something wrong, or are you the maintainer and want your own text here? Get in touch

Supports

Works locally Talks to your devices on your own network. No vendor cloud in between.
From the Marketplace Installed through the Marketplace, formerly HACS.
Actively maintained Its author released or pushed a change in the last six months.

Setting it up

Install it from the Marketplace and restart Home Assistant, then add it under Settings, Devices and services.

Profile Setup

Profile Name
Server Type

Server Configuration

Server URL
API Key
MCP Server Port
Gateway Host
Gateway Port
Gateway Token
Use SSL/TLS
Hermes Server URL

Model & Prompts

Model Name
System Prompt
Technical Instructions

Advanced Settings

Temperature
Max Response Tokens
Max History Messages
Max Tool Iterations
Control Home Assistant
Response Mode
Follow-up Phrases
End Conversation Words
Response Time Out
Debug Mode
Ollama Keep Alive
Ollama Context Window
Clean Responses for Voice Output
Session Key / Session IDOpenClaw session to use: a short key like 'main' or a full session ID like 'agent:main:main'.
Memory Session KeyStable key scoping Hermes long-term memory for this profile.

Shared MCP Server Settings

MCP Server Port
Web Search
Brave Search API Key
Additional Allowed IPs/Ranges
Enable Smart Entity Index
Max Entities Per Discovery (20-500)

After setup

Profile Name
Server URL
API Key
Model Name
System Prompt
Technical Instructions
Temperature
Max Response Tokens
Max History Messages
Max Tool Iterations
Control Home Assistant
Response Mode
Follow-up Phrases
End Conversation Words
Response Time Out
MCP Server Port
Web Search
Brave Search API Key
Additional Allowed IPs/Ranges
Enable Smart Entity Index
Max Entities Per Discovery (20-500)

Texts from the code of mike-nott/mcp-assist.

From the README

A Home Assistant conversation agent that uses MCP (Model Context Protocol) for efficient entity discovery, achieving 95% token reduction compared to traditional methods. Works with LM Studio, llama.cpp, Ollama, OpenAI, Google Gemini, Anthropic Claude, and OpenRouter.
✅ 95% Token Reduction: Uses MCP tools for dynamic entity discovery instead of sending all entities
✅ No Entity Dumps: Never sends 12,000+ token entity lists to the LLM
✅ Smart Entity Index: Pre-generated system structure index (~400-800 tokens) for context-aware queries
✅ Multi-Platform Support: Works with LM Studio, llama.cpp, Ollama, OpenAI, Google Gemini, Anthropic Claude, and OpenRouter
✅ Multilingual Support: 21 languages with localized UI, system prompts, and speech detection
✅ Multi-turn Conversations: Maintains conversation context and history
✅ Dynamic Discovery: Finds entities by area, type, device_class, state, or name on-demand
✅ Web Search Tools: Optional DuckDuckGo or Brave Search integration for current information
✅ Works with 1000+ Entities: Efficient even with large Home Assistant installations
✅ Multi-Profile Support: Run multiple conversation agents with different models

From the README of mike-nott/mcp-assist, MIT License. Shortened.