TcpSocket

You gain direct, low-level access to raw TCP sockets through this MCP server, letting you interact with network services and devices without HTTP wrappers. It supports multiple concurrent connections, buffering, and automatic responses based on trigger patterns, making it ideal for protocol debugging, legacy integrations, and automated testing of TCP-based systems.

How to use

You connect to a TCP service, send data, and read back responses using a sequence of simple operations. Start a connection, transmit the data you need, and then read from the connection buffer when data arrives. You can also set automatic responses that trigger when incoming data matches a pattern. When you are done, close the connection to free up resources.

How to install

Prerequisites: you need Python and a modern Python package manager. You can also use an offering that installs a runtime convenience tool for you.

Step 1: Clone the project repository and move into its directory.

Step 2: Install in editable mode using the recommended runtime helper.

Step 3: Run the server directly to start listening for TCP connections.

Configuration and startup details

You can configure the MCP server for Claude Desktop in two ways. The first approach uses an installed package directly, while the second runs from source.

Option 1: Using the installed package (recommended)

{
  "mcpServers": {
    "tcp-socket": {
      "command": "TcpSocketMCP",
      "env": {}
    }
  }
}

Usage from source

{
  "mcpServers": {
    "tcp-socket": {
      "command": "python",
      "args": ["/path/to/tcp-socket-mcp/run.py"],
      "env": {}
    }
  }
}

Available tools

tcp_connect

Opens a TCP connection to a host:port and returns a connection_id for subsequent operations.

tcp_send

Sends data over an established connection with optional encoding (utf-8, hex, base64) and an optional terminator.

tcp_read_buffer

Reads data from the connection’s receive buffer; supports partial reads via index/count and different formats (utf-8, hex, base64).

tcp_disconnect

Closes the connection and frees resources; automatically removes any associated triggers.

tcp_set_trigger

Registers automatic responses for pattern matches, including regex patterns with capture groups; triggers fire when data matches.

tcp_connect_and_send

Connects and sends in one atomic operation for time-sensitive handshakes.

tcp_list_connections

Lists all active connections with statistics.

tcp_connection_info

Provides detailed information about a specific connection.

tcp_buffer_info

Reports statistics about the receive buffer without consuming data.

tcp_clear_buffer

Clears the receive buffer for a connection.

tcp_remove_trigger

Removes a previously registered trigger.