shellagent

This MCP server provides an interface for shellagent-mcp, enabling you to run, discover, and call tools from a local or remote MCP client. You can operate it in a local stdio setup for rapid development or run it remotely over HTTP/SSE to connect from any compatible client. It supports listing available tools and invoking them, so you can automate tasks and workflows from your preferred MCP client.

How to use

You can use this server with any MCP client to list available tools and execute them from your environment. Choose a local stdio setup for fast iteration, or run a remote server and access it via Streamable HTTP or SSE endpoints. When you start the server, you can connect a client, request the list of tools, and call a specific tool with arguments. This lets you compose automated workflows and integrate the server into your tooling stack.

Key workflows you can perform include:

• Start a local session and list tools to see what you can call
• Call a tool by name with the required arguments and capture its result
• Run the server remotely and connect via a client using the SSE or Streamable HTTP endpoints
• Experiment with in-memory transports for in-process testing without starting a separate process

How to install

Prerequisites you need before installing or running the MCP server:

• Node.js 18 or newer must be installed on your machine
• A compatible MCP client such as VS Code, Cursor, Windsurf, Claude Desktop, or any MCP client

Local (Stdio) installation and configuration You configure the MCP server to run locally using stdio so the client communicates via in-process transport.

{
  "mcpServers": {
    "shellagent-mcp": {
      "command": "npx",
      "args": [
        "mcp-server-shellagent-mcp@latest"
      ]
    }
  }
}

Remote (SSE / Streamable HTTP) installation and usage You can start the server to be accessible remotely. Use a port for the server and connect through either SSE or Streamable HTTP endpoints.

Start the remote server on a chosen port (e.g., 8089):

npx mcp-server-shellagent-mcp --port 8089

Then use one of the endpoints from your MCP client configuration. The Streamable HTTP endpoint is recommended, and the SSE endpoint is also provided.

Streamable HTTP endpoint:

http://127.0.0.1::8089/mcp

SSE endpoint:

http://127.0.0.1::8089/sse

Client configuration variant for SSE:

{
  "mcpServers": {
    "shellagent-mcp": {
      "url": "http://127.0.0.1::8089/sse"
    }
  }
}

Client configuration variant for Streamable HTTP (if supported by your client):

{
  "mcpServers": {
    "shellagent-mcp": {
      "type": "streamable-http",
      "url": "http://127.0.0.1::8089/mcp"
    }
  }
}

In-memory call You can also test in-process without starting the CLI by connecting a client to an in-memory transport pair and using the API directly. This is useful for automated tests or quick experiments.

Additional notes

Development server You can run a local development server to experiment with MCP interactions.

Development command:

npm run dev

Available tools

listTools

Lists the tools that are available on the MCP server so you can see what you can call.

callTool

Invokes a specific tool by name with provided arguments and returns the result.