# Local(stdio) MCP Server with Docker

This project is a minimal local MCP server that exposes a `run_command` tool over stdio.


## Setup

```bash
cd ~path-to-directory/local-mcp-with-docker
uv sync
```


Start the script using:

```bash
uv run main.py
```

Note: from MCP 2.0 onward, there will be no output in the console after the server starts. The process stays alive and waits for stdio messages from an MCP client. You can inspect it using the debugger in a new terminal session.

Start the inspector:

```bash
npx -y @modelcontextprotocol/inspector
```

Then in the inspector UI:

- Transport: `stdio`
- Command: `uv`
- Args: `run`, `python`, `main.py`
- Working directory: `~path-to-directory/local-mcp-with-docker`

The inspector should show the registered tool `run_command` and allow you to call it directly.


## Build the Docker image

```bash
docker build -t terminal_tool_docker .
```

This creates the local Docker image used by the Claude Desktop MCP config.

## Claude Desktop config

Add this to your Claude Desktop MCP config file:

```json
{
  "mcpServers": {
    "local-mcp-with-docker": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "--init",
        "-e",
        "DOCKER_CONTAINER=true",
        "-e",
        "WORKSPACE_DIR=/root/mcp/workspace",
        "-v",
        "path-to-directory/local-mcp-with-docker:/root/mcp/workspace",
        "terminal_tool_docker"
      ]
    }
  }
}
```

This mounts your workspace into the container and points the server at the correct working directory inside the container.

# Claude screenshot
![alt text](image.png)