Tabby-MCP-Server

Tabby-MCP-Server is an MCP server that integrates directly into the Tabby Terminal application as a plugin, exposing active terminal sessions through the Model Context Protocol. It connects local shells and SSH-backed terminal tabs to external AI clients such as Claude Code, Cursor, Windsurf, and OpenAI Codex CLI. Software engineers and system administrators use it to allow AI assistants to interact safely with command-line environments without leaving the primary terminal window. The server enables AI clients to discover available terminal tabs, execute shell commands, retrieve exit codes, inspect visible or historical terminal buffers, and page through lengthy command outputs. It also includes an optional pair programming mode that renders visual confirmation dialogs inside Tabby before running commands, giving users complete visibility and manual approval control over agent-driven actions.

Category: Cloud & Infrastructure

Tags: automation, shell, tabby, terminal

Visit Tabby-MCP-Server

How to install and configure Tabby-MCP-Server

1. Open Tabby Terminal, navigate to Settings → Plugins, and install the "Tabby MCP" plugin, then restart Tabby. 2. Alternatively, clone the repository and build it locally using Docker: bash git clone https://github.com/thuanpham582002/tabby-mcp-server.git cd tabby-mcp-server make build-dist bash scripts/copy_to_plugin_folder.sh 3. Restart Tabby and ensure at least one terminal tab is open. Verify the service is active with curl http://localhost:3001/health. 4. For Cursor or general SSE-capable clients, add the server to your configuration file: json { "mcpServers": { "tabby-mcp": { "type": "sse", "url": "http://localhost:3001/sse" } } } 5. For Claude Code, run: bash claude mcp add --transport sse tabby-mcp http://localhost:3001/sse

What you can do with Tabby-MCP-Server

• Running terminal diagnostics and build commands automatically from an AI assistant while viewing execution in Tabby. - Inspecting visible terminal buffers and terminal history to allow AI agents to debug failures. - Querying active local and SSH-backed terminal session lists to target specific running environments. - Paginating and retrieving long logs or large command output streams using dedicated output identifiers. - Safely executing developer commands with human-in-the-loop confirmation dialogs in pair programming mode.

Key facts

• https://github.com/thuanpham582002/tabby-mcp-server
• Cloud & Infrastructure, Developer Tools & Code Intelligence
• automation, shell, tabby, terminal

Part of MCP Servers

What is Tabby-MCP-Server?

Tabby-MCP-Server is a plugin for Tabby Terminal that implements the Model Context Protocol over Server-Sent Events. It allows MCP-compliant AI assistants to interact with terminal tabs, run commands, and read console output directly.

How do I install Tabby-MCP-Server?

You can install it directly inside Tabby by opening Settings, clicking Plugins, searching for Tabby MCP, and restarting the application. Alternatively, you can build the plugin locally using the provided Docker build commands and installation script.

Which MCP clients work with Tabby-MCP-Server?

Any AI client that supports SSE transport can connect to Tabby-MCP-Server. This includes Claude Code, Cursor, Windsurf, OpenAI Codex CLI, and any custom MCP client pointing to the local SSE endpoint at port 3001.

What tools does Tabby-MCP-Server provide?

It provides four core tools: get_ssh_session_list to discover open terminal tabs, exec_command to run commands in a specific tab, get_terminal_buffer to read scrollback history, and get_command_output to paginate through long execution results.

Is Tabby-MCP-Server open source?

Yes, Tabby-MCP-Server is open-source software licensed under the MIT License. The full source code, build scripts, and documentation are publicly accessible on GitHub.