MCP Sinstaller

MCP安全安装程序(mcp-sinstaller)

该项目提供了一个 MCP服务器安装程序 它可以自动安装和容器化GitHub存储库中的其他MCP服务器。它使用模型上下文协议（MCP）采样功能来分析存储库并创建适当的docker映像。

https://github.com/user-attachments/assets/6ad17a8b-c88f-4cba-abea-3a1f5288ed3c

需求

• 码头工人
• 启用采样的MCP客户端

在VS代码内部人员中启用采样（推荐）

将以下内容添加到VS代码设置中：

"chat.mcp.serverSampling": {
    "Global in Code - Insiders: sampling": {
        "allowedDuringChat": true
    }
}

用法

以VS代码运行服务器（推荐）

1. uvx

首先，克隆存储库：

git clone https://github.com/semcp/mcp-sinstaller.git

然后，将以下内容添加到您的VS代码设置中：

"mcp-sinstaller": {
    "type": "stdio",
    "command": "uvx",
    "args": [
            "--from",
            "/path/to/mcp-sinstaller",
            "python",
            "/path/to/mcp-sinstaller/mcp_installer.py",
    ]
},

服务器命名和多个服务器

安装程序使用结构化命名方案： owner-repo-servername 以避免从同一存储库安装多个服务器时发生冲突。

示例：

• https://github.com/modelcontextprotocol/servers/tree/main/src/fetch → modelcontextprotocol-servers-fetch
• https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem → modelcontextprotocol-servers-filesystem
• https://github.com/github/github-mcp-server → github-github-mcp-server

这允许您从存储库安装多个MCP服务器，例如 modelcontextprotocol/servers 没有命名冲突。

示例（GitHub MCP服务器）

提示：“安装https://github.com/github/github-mcp-server"

GitHub Copilot:
I'll install the GitHub MCP server from the repository you specified.

sampling (MCP Server):

Input:

{
  "url": "https://github.com/github/github-mcp-server"
}

Output:

Successfully built Docker image for MCP server 'github-mcp-server'!

Repository: https://github.com/github/github-mcp-server
Location: /Users/mossaka/.mcp-installer/servers/github-mcp-server

DOCKER IMAGE:
- Using existing Dockerfile
- Image: mcp-installer-github-mcp-server
- Size & Created: REPOSITORY:TAG                           SIZE      CREATED AT
mcp-installer-github-mcp-server:latest   59.5MB    2025-06-09 14:52:46 -0700 PDT
- Ports: {}
- Environment Variables: ['GITHUB_PERSONAL_ACCESS_TOKEN', 'GITHUB_TOOLSETS', 'GITHUB_DYNAMIC_TOOLSETS', 'GITHUB_READ_ONLY', 'GITHUB_HOST']
- Required Secrets: ['GITHUB_PERSONAL_ACCESS_TOKEN']
- Metadata: /Users/mossaka/.mcp-installer/servers/github-mcp-server/metadata.json

TO RUN THE CONTAINER (STDIO Transport):

1. Set required environment variables (if any):
export GITHUB_PERSONAL_ACCESS_TOKEN=your_actual_github_personal_access_token_here

2. Run the container interactively:
   docker run -i --rm --name mcp-github-mcp-server-container -e GITHUB_PERSONAL_ACCESS_TOKEN=GitHub personal access token (required) -e GITHUB_TOOLSETS=Comma-separated list of toolsets to enable (optional) -e GITHUB_DYNAMIC_TOOLSETS=Set to 1 to enable dynamic tool discovery (optional) -e GITHUB_READ_ONLY=Set to 1 to enable read-only mode (optional) -e GITHUB_HOST=Custom GitHub API endpoint for GitHub Enterprise (optional) -e GITHUB_PERSONAL_ACCESS_TOKEN=${GITHUB_PERSONAL_ACCESS_TOKEN} mcp-installer-github-mcp-server
   
   Note: This server uses STDIO transport (stdin/stdout communication).
   - The container runs interactively (-i flag)
   - It will be automatically removed when it exits (--rm flag)
   - No port mapping needed as it uses standard input/output

3. To stop the container:
   Ctrl+C in the terminal where it's running

4. For background operation with MCP client:
   Use your MCP client to connect via subprocess/stdio

Great! The GitHub MCP server has been successfully installed. The installation completed successfully and built a Docker image.

Here's what was installed:

Server Name: github-mcp-server
Repository: https://github.com/github/github-mcp-server
Docker Image: mcp-installer-github-mcp-server:latest (59.5MB)
Transport Type: STDIO (communicates via standard input/output)
Required Configuration: The server requires a GITHUB_PERSONAL_ACCESS_TOKEN environment variable to authenticate with GitHub's API.

Optional Environment Variables:

GITHUB_TOOLSETS: Comma-separated list of toolsets to enable
GITHUB_DYNAMIC_TOOLSETS: Set to 1 to enable dynamic tool discovery
GITHUB_READ_ONLY: Set to 1 to enable read-only mode
GITHUB_HOST: Custom GitHub API endpoint for GitHub Enterprise
The server is now ready to use. To run it, you'll need to set your GitHub personal access token and then run the Docker container with the STDIO transport for communication with MCP clients.

之后，你就可以跑了 docker 在代理的MCP设置中运行容器。

VS代码

"github": {
    "type": "stdio",
    "command": "docker",
    "args": [
        "run",
        "-i",
        "--rm",
        "-e",
        "GITHUB_PERSONAL_ACCESS_TOKEN=
",
        "mcp-installer-github-mcp-server",
    ]
}

示例（获取服务器）

提示：“安装https://github.com/modelcontextprotocol/servers/tree/typescript-servers-0.6.2/src/fetch"

I'll install the fetch MCP server from the specific path you provided.

sampling (MCP Server)

Input:
{
  "url": "https://github.com/modelcontextprotocol/servers/tree/typescript-servers-0.6.2/src/fetch"
}

Output:
Successfully built Docker image for MCP server 'modelcontextprotocol-servers-fetch'!

Repository: https://github.com/modelcontextprotocol/servers/tree/typescript-servers-0.6.2/src/fetch
Location: /Users/mossaka/.mcp-sinstaller/servers/modelcontextprotocol-servers-fetch

DOCKER IMAGE:
- Generated new Dockerfile
- Image: mcp-sinstaller-modelcontextprotocol-servers-fetch
- Size & Created: REPOSITORY:TAG                                             SIZE      CREATED AT
mcp-sinstaller-modelcontextprotocol-servers-fetch:latest   654MB     2025-06-09 16:28:35 -0700 PDT
- Ports: {}
- Environment Variables: []
- Required Secrets: []
- Metadata: /Users/mossaka/.mcp-sinstaller/servers/modelcontextprotocol-servers-fetch/metadata.json

TO RUN THE CONTAINER (STDIO Transport):

1. Set required environment variables (if any):
   No secrets required

2. Run the container interactively:
   docker run -i --rm --name mcp-modelcontextprotocol-servers-fetch-container mcp-sinstaller-modelcontextprotocol-servers-fetch
   
   Note: This server uses STDIO transport (stdin/stdout communication).
   - The container runs interactively (-i flag)
   - It will be automatically removed when it exits (--rm flag)
   - No port mapping needed as it uses standard input/output

3. Use your MCP client to connect via subprocess/stdio transport

✅ Fetch MCP Server installed successfully!

The server has been built and is ready to use. Here are the key details:

Server Information:

Name: modelcontextprotocol-servers-fetch
Repository: https://github.com/modelcontextprotocol/servers/tree/typescript-servers-0.6.2/src/fetch
Transport: STDIO (standard input/output)
Image Size: 654MB

VS代码

"fetch": {
    "type": "stdio",
    "command": "docker",
    "args": [
        "run",
        "-i",
        "--rm",
        "mcp-sinstaller-modelcontextprotocol-servers-fetch"
    ]
},