Local AI integration

Let your AI tools help at home.

Chantico includes a local Model Context Protocol server, so compatible agentic applications can discover devices, run scenes, and manage automations—with only the permissions you choose.

Connect an agent Get Chantico

Chantico connected locally

Set the studio lights to 40% and turn off the hallway.

Done. I adjusted the studio lights and turned off the hallway.

Device control allowed by you

A local bridge, not another cloud account.

MCP is an open standard that lets AI applications use clearly defined tools. Chantico runs the server on your Windows PC. Desktop clients can connect over local standard input; supported hosted clients can use an outbound secure tunnel that you configure explicitly.

1Your agentClaude or another MCP client

2Chantico MCPPermission-checked local tools

3Your homeMatter and Philips Hue

Useful capabilities, deliberately scoped.

Understand

Check the controller, list devices, and read a device's current state.

list_devices · get_device_state

Control

Turn devices on or off, adjust supported controls, and run scenes.

control_device · run_scene

Automate

List automations and enable or disable the ones you select.

list_automations · enable_automation

Connect your AI tool.

Choose your client to see its setup instructions.

Start in Chantico

First, open Chantico Settings → AI & developer tools, enable external tools, and choose exactly which capability groups the agent may use. Use Copy MCP configuration there to get the correct installed path, then paste it into your client in place of the example path below.

Find your chantico.exe path with PowerShell

For a Microsoft Store installation, open Windows PowerShell on the Chantico PC under your usual Windows account and paste this script. No administrator access is needed. It only reads the installation location and checks that the CLI exists.

$chanticoPackage = Get-AppxPackage -Name '44060saulricardo.chantico' |
    Sort-Object Version -Descending | Select-Object -First 1
if (-not $chanticoPackage) {
    throw 'Chantico was not found for this Windows account. Install it from the Microsoft Store, then try again.'
}
$chanticoExe = Join-Path $chanticoPackage.InstallLocation 'Cli\chantico.exe'
if (-not (Test-Path -LiteralPath $chanticoExe -PathType Leaf)) {
    throw 'The Chantico CLI was not found. Update Chantico, or use Copy MCP configuration in its Settings.'
}
'Full path (for terminal commands):'
$chanticoExe
'JSON value (including quotes, for configuration files):'
ConvertTo-Json -InputObject $chanticoExe

For terminal commands, replace the example path inside the quotes with the full path printed above. For JSON configurations, replace the entire quoted example value with the JSON value printed above—it already escapes the backslashes.

Store updates can change the installation path. If a connection stops working after an update, run this again and update your client configuration. For other installations, use Copy MCP configuration in Chantico instead.

Claude Desktop Claude Code ChatGPT VS Code · Copilot Visual Studio · Copilot Copilot CLI Other MCP clients

Add to your configuration

Open Developer settings, edit the MCP configuration, and add:

{
  "mcpServers": {
    "chantico": {
      "command": "C:\\...\\Chantico\\Cli\\chantico.exe",
      "args": ["mcp", "serve"]
    }
  }
}

Restart Claude Desktop or open a new chat so it refreshes the tool list.

Register from a terminal

Run these commands once for your Windows account:

claude mcp add chantico --scope user -- "C:\...\Chantico\Cli\chantico.exe" mcp serve
claude mcp get chantico

The second command confirms that the server is registered.

Connect through Secure MCP Tunnel

ChatGPT cannot launch Chantico's local server directly. Create a tunnel in the OpenAI Platform, download tunnel-client, then initialize it with Chantico as the local stdio command:

tunnel-client init --sample sample_mcp_stdio_local --profile chantico --tunnel-id <YOUR_TUNNEL_ID> --mcp-command "C:\...\Chantico\Cli\chantico.exe mcp serve"

tunnel-client doctor --profile chantico --explain
tunnel-client run --profile chantico

Keep tunnel-client run --profile chantico running on the Chantico PC.
In ChatGPT, enable Developer mode and go to Settings → Apps → Create.
Choose Tunnel, select your tunnel, scan the available tools, and create the app.
Open a new chat and select Chantico from the tools menu before asking it to control your home.

Requires a supported ChatGPT plan or workspace and an OpenAI Platform tunnel. Requests travel outbound through the secure tunnel; Chantico is not exposed through an inbound public port.

Connect Copilot in VS Code

Run MCP: Open User Configuration from the Command Palette and add:

{
  "servers": {
    "chantico": {
      "type": "stdio",
      "command": "C:\\...\\Chantico\\Cli\\chantico.exe",
      "args": ["mcp", "serve"]
    }
  }
}

Open Copilot Chat, switch to Agent mode, then use Configure Tools to enable the Chantico tools you want available.

Connect Copilot in Visual Studio

In Visual Studio 2026, open or create %USERPROFILE%\.mcp.json. Add Chantico to its servers section, keeping any existing servers:

{
  "servers": {
    "chantico": {
      "type": "stdio",
      "command": "C:\\...\\Chantico\\Cli\\chantico.exe",
      "args": ["mcp", "serve"]
    }
  }
}

Save the file, open Copilot Chat in Agent mode, then select Tools and enable the Chantico tools you want available.

Visual Studio MCP setup guide →

Connect Copilot from your terminal

With GitHub Copilot CLI installed, run these commands in a terminal on your Chantico PC. Replace the example path with your installed Chantico path:

copilot mcp add chantico -- "C:\...\Chantico\Cli\chantico.exe" mcp serve
copilot mcp get chantico

The second command confirms registration. Start Copilot CLI and ask it to use Chantico; review its tool requests before approving them.

Copilot CLI MCP setup guide →

Use the standard local transport

For clients that accept a generic stdio server definition, use:

{
  "type": "stdio",
  "command": "C:\\...\\Chantico\\Cli\\chantico.exe",
  "args": ["mcp", "serve"]
}

The client must run on the same Windows account and support local stdio MCP servers.

Speak naturally. The agent chooses the right Chantico tool.

“Which lights are still on downstairs?”

“Set the studio light to 40%.”

“Run my Good night scene.”

“Disable the weekday morning automation.”

Private by design.

MCP access is off by default. For local clients, Chantico uses standard input and a current-user Windows named pipe; it does not open an inbound network port. Agents receive no arbitrary command-line access, and every request is checked against your current settings.

You can independently allow device visibility, device control, scenes, or automation management. Changes can require a temporary approval window granted inside Chantico, while read-only tools remain available.

Chantico keeps a bounded MCP activity history locally on your PC. It records the tool, outcome, time, and opaque target ID—not prompts, conversations, command values, device names, or room names. You can clear it from Settings at any time.

Read the privacy policy →