Installation

System requirements

Python (version >= 3.10).
lys that supports --remote option. lys and lys_mcp can be installed in different environments.
Claude Code (https://claude.com/claude-code), if you want to control lys from Claude.

Installation from source

Update pip:
```
pip install --upgrade pip
```
Clone lys_mcp. If you do not have git, you can download the source code from GitHub (https://github.com/lys-devel/lys_mcp):
```
git clone https://github.com/lys-devel/lys_mcp.git
```
Install lys_mcp by pip. The [mcp] option installs the MCP server (lys-mcp command) together with the command line client (lys-remote command):
```
cd lys_mcp
pip install -e ".[mcp]"
```
If you only need the command line client, [mcp] can be omitted:
```
pip install -e .
```
Make sure that the commands are installed:
```
lys-remote --help
which lys-mcp
```

Registration to Claude Code

Claude Code needs to know the MCP server. Register it once by the command below:

claude mcp add --scope user lys -- lys-mcp

--scope user enables lys tools in all directories. Restart Claude Code after the registration, and confirm that lys is listed by /mcp command in Claude Code.

If you want to enable lys tools only in a specific project, put .mcp.json in the project directory instead:

{
  "mcpServers": {
    "lys": {
      "command": "lys-mcp"
    }
  }
}

If lys-mcp is installed in a virtual environment, use the full path of the command (e.g. ~/.venv/lys/bin/lys-mcp).