Tabby MCP, developed by Gentlemanhu, is an MCP server designed to link AI coding agents and local AI assistants with the Tabby Terminal application. Built with TypeScript and running on Node.js, this free plugin exposes active terminal sessions through the Model Context Protocol. It allows AI clients to interact with the terminal environment, executing commands and retrieving output. The tool supports various workflows for developers and automation, making terminal operations accessible to external programs.

What does Tabby MCP connect to?

Tabby MCP functions as a specialized plugin for the Tabby Terminal application, connecting directly to its active terminal sessions. It exposes these sessions through the Model Context Protocol (MCP), serving as an MCP server. This connection allows external clients, particularly AI coding agents and local AI assistants, to interact with the terminal environment. The primary transport protocol for client connections is Server-Sent Events (SSE), with clients typically connecting to an SSE endpoint such as "http://localhost:3001/sse" for communication. Compatible AI clients include Claude Code, Cursor, Windsurf, and Codex / OpenAI Codex CLI.

Core features and tools

Tabby MCP exposes several callable tools for interacting with terminal sessions. These include get_ssh_session_list, which lists available Tabby terminal sessions, and exec_command, for executing commands in a selected terminal tab. Developers can also use get_terminal_buffer to retrieve current terminal content and get_command_output to obtain complete command results. While these four are explicitly named, the system is noted to include additional MCP tools, with one source mentioning up to 36 tools, including SFTP support, expanding its potential utility.

Who is it for?

This program is intended for developers, AI coding agents, local AI assistants, and those building custom MCP clients. It enables AI clients to run shell commands in live Tabby terminal sessions, providing a direct interface for automated operations. Developers can also retrieve full command output for debugging or analysis, enhancing their workflow. The tool supports automating repetitive terminal workflows through an MCP-compatible assistant. Skip this if you do not use Tabby Terminal or require direct terminal interaction from AI clients.

How to get started

No credentials are required for basic operation, though an optional "Pair Programming Mode" can be enabled for command confirmation. Clients primarily connect using the Server-Sent Events (SSE) transport protocol.

Use cases

• An AI client running shell commands in a live Tabby terminal session.
• A developer retrieving full command output for debugging or analysis.
• Automating repetitive terminal workflows through an MCP-compatible assistant.

Tabby MCP offers a direct way for AI agents and developers to interact with Tabby Terminal sessions. Its SSE transport and defined tools like exec_command enable automation and data retrieval. The optional Pair Programming Mode adds a layer of safety for command execution. This tool is a valuable addition for those integrating AI with terminal environments. Start by using exec_command to run a shell command in a live Tabby terminal session.