tcpsh-mcp

tcpsh-mcp

MCP server for interactive TCP connection management, enabling opening listeners, managing sessions, port forwarding, proxy with logging, and local command execution from any MCP host.

Category
Visit Server

README

tcpsh-mcp — MCP Server for TCP Connection Management

MCP server for tcpsh — the interactive TCP connection manager.

MCP server exposing 13 tools for interactive TCP connection management: open listeners, interact with sessions, forward ports, proxy traffic, and run local commands — all from any MCP-compatible AI host (Claude Desktop, VS Code Copilot, etc.).


Installation

This server is installed on demand from GitHub — no npm registry required.

Add to your MCP host configuration:

{
  "mcpServers": {
    "tcpsh": {
      "command": "npx",
      "args": ["-y", "github:nchgroup/tcpsh-mcp"]
    }
  }
}

npx -y github:nchgroup/tcpsh-mcp clones the repository, runs npm install, and executes the bin entry — no manual setup required.

Requirements

  • Node.js ≥ 18.0.0
  • Internet access on first run (GitHub clone + npm install)

Tools

# Tool Description
1 open_port Open a TCP listener
2 close_port Close a listener and its sessions
3 list_ports List open ports and session counts
4 list_sessions List active sessions with traffic counters
5 send_to_session Send data to a TCP session
6 read_from_session Read buffered data from a TCP session
7 kill_session Terminate a session (FIN or RST)
8 session_info Detailed info about a session
9 add_forward Start a transparent TCP forward
10 add_proxy Start a TCP proxy with traffic logging
11 list_forwards List active forwards and proxies
12 remove_forward Stop a forward or proxy
13 exec_local Run a shell command on the local machine

Tool Reference

open_port

port   (number, required)  TCP port to listen on
host   (string, optional)  Bind address or interface name — default: 0.0.0.0

host accepts an IP address, a network interface name, or can be omitted:

open_port(4444)                  # bind 0.0.0.0:4444
open_port(4444, "127.0.0.1")     # loopback only
open_port(4444, "tun0")          # resolves tun0 → its IPv4 address
open_port(443,  "eth0")          # resolves eth0 → its IPv4 address

In standalone mode (no TCPSH_SERVER) the interface is resolved locally via os.networkInterfaces().
In remote mode the interface name is sent as-is to the tcpsh server, which resolves it there.

close_port

port   (number, required)  Port to close (closes all its sessions too)

send_to_session

port   (number, required)  Port the session belongs to
idx    (number, optional)  Session index, 1-based — default: 1
data   (string, required)  Raw data to send (no newline added automatically)

read_from_session

port   (number, required)  Port the session belongs to
idx    (number, optional)  Session index — default: 1

Drains the internal RX buffer. Call again to receive subsequent data.

kill_session

port   (number, required)  Port the session belongs to
idx    (number, optional)  Session index — default: 1
force  (boolean, optional) true = RST (immediate), false = FIN (graceful) — default: false

add_forward

local_port   (number, required)  Local port to listen on
remote_host  (string, required)  Destination host
remote_port  (number, required)  Destination port

add_proxy

local_port   (number, required)  Local port to listen on
remote_host  (string, required)  Destination host
remote_port  (number, required)  Destination port
log_file     (string, optional)  Path to write hex traffic log

exec_local

command  (string, required)  Shell command executed via sh -c

Example Prompts

Open a reverse shell catcher:

"Open port 4444 and wait for a connection"

Read shell output:

"Read the output from the session on port 4444"

Send a command to a session:

"Send the command 'id\n' to port 4444"

Set up port forwarding:

"Forward local port 8080 to 10.0.0.1:80"

Intercept HTTP traffic:

"Start a proxy on port 8080 pointing to 10.0.0.1:80, log to /tmp/http.log"

Run a local network scan:

"Run nmap -sV localhost and return the output"


Remote Mode

tcpsh-mcp can delegate all TCP management to a remote tcpsh server instead of running a local in-process manager. This is useful when:

  • the AIhost is sandboxed and cannot open TCP ports directly
  • you want state to persist even if the MCP process restarts
  • multiple AI sessions should share the same network state

Prerequisites

Start a tcpsh server on the target machine:

tcpsh -server 0.0.0.0:9000

The token printed at startup encrypts all traffic with ChaCha20-Poly1305.

Enable remote mode

Pass two environment variables when starting the MCP server. Example MCP host configuration:

{
  "mcpServers": {
    "tcpsh-remote": {
      "command": "npx",
      "args": ["-y", "github:nchgroup/tcpsh-mcp"],
      "env": {
        "TCPSH_SERVER": "127.0.0.1:9000",
        "TCPSH_TOKEN": "aBcDeFgHiJkLmNoPqRsTuVwXyZ012345"
      }
    }
  }
}

Or on the command line:

TCPSH_SERVER=127.0.0.1:9000 TCPSH_TOKEN=<TOKEN> node src/index.js

What works remotely

Tool Remote support
open_portremove_forward (12 tools) ✅ Delegated to tcpsh server
exec_local ⚠️ Always runs locally (on the MCP server host)

When neither TCPSH_SERVER nor TCPSH_TOKEN is set, tcpsh-mcp falls back to its built-in in-process TCP manager (default behaviour).


Architecture

tcpsh-mcp/
├── package.json            bin entry + @modelcontextprotocol/sdk dep
└── src/
    ├── index.js            McpServer + StdioServerTransport + tool registration
    │                       (selects TcpManager or RemoteTcpManager at startup)
    ├── tcp-manager.js      Local TCP state: listeners, sessions, RX buffers, forwards, proxies
    ├── remote-manager.js   Remote mode: encrypted ChaCha20-Poly1305 frames over TCP
    └── tools.js            13 tool definitions (inputSchema + handlers)
  • Transport: stdio — compatible with all MCP hosts
  • State: in-process (default) or on the remote tcpsh server (remote mode)
  • RX buffer: received data is buffered per session and drained on read_from_session
  • SDK: @modelcontextprotocol/sdk v1.x (stable)

Manual Testing

Send raw JSON-RPC to the server's stdin:

# Start the server
node src/index.js

# In another terminal / via pipe:
printf '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}\n{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}\n' \
  | node src/index.js

Related Projects

  • tcpsh — Go implementation (REPL binary) — the project this MCP wraps
  • tcpsh-mcp — this repository

License

MIT

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured