ha-mcp
A Model Context Protocol (MCP) server that provides AI assistants with access to Home Assistant, enabling smart home control and automation management.
ha-mcp
A Model Context Protocol (MCP) server that provides AI assistants with access to Home Assistant, enabling smart home control and automation management.
Features
- 38 Specialized Tools: Entity queries, automation CRUD, helper management, scripts, scenes, devices, areas, labels, floors, zones, persons, tags, traces, blueprints, updates, todos, calendars, cameras, dashboards, and more
- Hybrid Architecture: WebSocket for most operations, REST API for automation/script/scene CRUD
- Complete CRUD: Create, read, update, delete automations/scripts/scenes/helpers
- Deep System Access: Query registries, analyze dependencies, access logbook, validate config
- Flexible Output: Natural language (LLM-optimized) and JSON formats
- Access Control: Read-only mode, whitelist/blacklist, fine-grained action-level control
- Auto-Reconnect: Automatic reconnection with exponential backoff
- Post-Mutation Confirmation: Automatic state polling after create/update/delete confirms changes
vs. Other MCP Servers for Home Assistant
Two alternatives exist: the official HA MCP integration (built-in, ~10 intent-based tools) and the community homeassistant-ai/ha-mcp (Python/FastMCP, 95+ tools).
Choose ha-mcp if you need:
- Full automation/script/scene/helper lifecycle management (create, edit, delete)
- Advanced analysis (dependencies, cross-references, automation coverage)
- System administration (registry queries, config validation, logbook, history)
- Media management (browser, camera streams), HACS, and dashboard access
- Reliable LLM tool selection — 38 consolidated tools reduce selection errors compared to 95+ fine-grained alternatives
Choose the official integration if you need entity-level security or no external infrastructure.
See docs/feature-comparison.md for a detailed three-way feature matrix.
Installation
From Binary
Download the latest release from the Releases page.
# Linux/macOS
tar -xzf ha-mcp_linux_amd64.tar.gz
chmod +x ha-mcp
sudo mv ha-mcp /usr/local/bin/
# Windows: extract ha-mcp_windows_amd64.zip and add to PATH
From Source
Requires Go 1.26 or later.
git clone https://github.com/zorak1103/ha-mcp.git
cd ha-mcp
go build -o ha-mcp ./cmd/ha-mcp
Linux Packages
RPM and DEB packages are available in the releases:
sudo dpkg -i ha-mcp_amd64.deb # Debian/Ubuntu
sudo rpm -i ha-mcp_amd64.rpm # RHEL/Fedora
Docker
docker pull zorak1103/ha-mcp:latest
docker run -d --name ha-mcp -p 8080:8080 \
-e HA_URL=http://homeassistant.local:8123 \
zorak1103/ha-mcp:latest
See docs/configuration.md for Docker options, HTTPS/WSS, proxy support, and all environment variables.
Quick Start
-
Get a long-lived access token from your Home Assistant profile page.
-
Start the server:
# With flags
ha-mcp --ha-url http://homeassistant.local:8123 --ha-token your-token
# Or initialize config files first
ha-mcp init # creates config.yaml and .env
ha-mcp # start with config file
- Connect your AI client. Example for Claude Desktop:
{
"mcpServers": {
"homeassistant": {
"type": "http",
"url": "http://localhost:8080",
"headers": { "Authorization": "Bearer your-ha-access-token" }
}
}
}
See docs/configuration.md for Cline, opencode, and other client configurations.
Available Commands
| Command | Description |
|---|---|
ha-mcp | Start the MCP server |
ha-mcp init | Create config.yaml and .env in current directory |
ha-mcp config | Display effective configuration (tokens masked) |
ha-mcp --help | Show help and available flags |
Available Tools
38 tools organized by domain. Full reference at docs/tools.md.
| Category | Count | Highlights |
|---|---|---|
| Entity | 5 | query_entities (history/stats/health), get_state, analyze_entity |
| Registry | 10 | get_registry, manage_area/label/floor/zone/person/tag/entity/device |
| Automation | 1 | manage_automation (CRUD, toggle, coverage, JSON Patch + semantic patch) |
| Helpers | 2 | manage_helper (26 types), helper_action |
| Scripts & Scenes | 2 | manage_script, manage_scene (CRUD + execute/activate + JSON Patch + semantic patch) |
| Analysis | 3 | analyze_entity, get_entity_dependencies, analyze_target |
| Services | 2 | call_service, list_services |
| History/Logbook | 2 | query_entities modes, get_logbook (entries + correlation) |
| Dashboards/Media | 4 | manage_dashboard (JSON Patch + semantic patch), browse_media, manage_camera, sign_media_path |
| Calendars & Todos | 2 | manage_calendar, manage_todo |
| System/Admin | 7 | get_system_info, validate_config, manage_update, manage_blueprint |
| HACS | 1 | manage_hacs (list, download, install, custom repos) |
Access Control
ha-mcp provides read-only mode, whitelist, and blacklist filtering at the tool and action level:
# config.yaml — read-only monitoring
server:
read_only: true
# Or block specific operations
server:
tool_filter:
blacklist:
- "call_service"
- "manage_*:delete"
See docs/access-control.md for glob patterns, category filtering (*:write), and example scenarios.
Architecture
AI Client → HTTP/JSON-RPC → ha-mcp MCP Server
│
┌────────────────────┴────────────────────┐
│ WebSocket (primary) │ REST API
│ - State queries, service calls │ - Automation CRUD
│ - Helper CRUD, Registry access │ - Script/Scene CRUD
└────────────────────┬────────────────────┘
│
Home Assistant
See docs/architecture.md for project structure, build commands, and integration test setup.
Troubleshooting
See docs/troubleshooting.md for WebSocket connection issues, debug mode, and common error solutions.
Development
Prerequisites: Go 1.26+, golangci-lint v2, Docker (optional)
go build -o ha-mcp ./cmd/ha-mcp # Build
go test ./... # Unit tests
golangci-lint run --timeout=5m ./... # Lint
See docs/architecture.md for integration test setup and docs/integration-tests.md for the full test suite documentation.
Contributing
- Fork the repository and create a feature branch
- Make your changes with tests
- Ensure CI passes:
golangci-lint run ./... go test -race ./... - Open a Pull Request
Pull Request Guidelines
- Ensure CI checks pass (lint, test, security scans)
- Update documentation if needed
- Add tests for new functionality
- Keep commits focused and atomic
License
GPL-3.0 License - see LICENSE for details.
Acknowledgments
- Model Context Protocol specification
- Home Assistant WebSocket API
- coder/websocket - Pure Go WebSocket library
相關伺服器
MCP Media Processing Server
A server for media processing, offering powerful video and image manipulation using FFmpeg and ImageMagick.
LinkedIn Ads MCP
Connect LinkedIn Ads to Claude or ChatGPT via Two Minute Reports MCP to get clear insights into campaign performance, impressions, CTR, CPC, leads, and conversions.
Hyteria MCP
A server for looking up the daily menu at the Hyteria (B1) restaurant.
SEOforGPT MCP
Track brand visibility across ChatGPT, Claude, and Perplexity — monitor AI mentions, analyze competitors, and generate content optimized for AI citations.
OP.GG
Access real-time gaming data across popular titles like League of Legends, TFT, and Valorant, offering champion analytics, esports schedules, meta compositions, and character statistics.
Cinderfi
Tax-aware retirement planning for Canada and the US — CPP/OAS, Social Security, RRSP/TFSA/401k/IRA, Monte Carlo, withdrawal optimization.
HomeMCPBridge
Native macOS HomeKit integration for AI assistants via Model Context Protocol
LLM Brand Monitor MCP Server
MCP server for LLM Brand Monitor — track how AI models mention your brand. Requires an API key (LBM_API_KEY) from llmbrandmonitor.com.
Pepesto MCP
Pepesto gives your agent the ability to turn any recipe (a URL, plain text, or a photo) into a matched basket of real supermarket products with live prices, across 26 European supermarkets. The MCP covers the recipe → matched cart half of the grocery shopping workflow (parse / search / map ingredients to SKUs / check catalogs)
FHIR MCP Server
FHIR MCP Server – helping you expose any FHIR Server or API as a MCP Server.