Terminal MCP Server
Let your coding agent drive TUIs and CLIs like a browser—inspect screens, click menus, and automate terminal apps in headless CI or visual tmux mode.
Overview
Terminal MCP Server is a MCP server for the Ship phase that lets AI agents inspect and control terminal UIs through a structured state tree and ten automation tools.
What is this MCP server?
- 10 MCP tools for spawn, input, snapshot, and structured terminal interaction
- Terminal State Tree (TST) with detected menus, tables, buttons, and inputs
- Headless mode for CI plus visual mode with xterm/tmux on Linux, macOS, and WSL
- Multi-platform PTY management on x64 and arm64
- Playwright-style workflows for TUI apps and CLI tools
- 10 MCP tools for terminal interaction
- Supports headless CI mode and visual xterm/tmux mode
- Platforms: linux, macos, wsl; architectures x64 and arm64
What problem does it solve?
Raw terminal output is opaque to agents, so testing TUIs and guiding users through CLI wizards breaks down into fragile string matching and manual tmux sessions.
Who is it for?
Indie builders shipping TUI dashboards, installers, or internal CLIs who want agent-driven regression tests before each release.
Skip if: Teams that only need one-off shell commands with no interactive prompts, or pure web apps with no terminal UI.
What do I get? / Deliverables
After you register the server, your agent can spawn PTYs, read structured UI elements, and run visual or headless terminal test workflows on Linux, macOS, or WSL.
- Agent-driven terminal sessions with structured UI snapshots
- Repeatable headless or visual TUI test workflows
- Cross-platform PTY automation on x64/arm64
Recommended MCP Servers
Journey fit
Terminal automation sits on the Ship shelf because solo builders use it to programmatically test and regression-check CLI/TUI products before release. Testing is the canonical subphase: the server exposes a Terminal State Tree and UI element detection so agents can assert behavior instead of only running shell one-liners.
How it compares
MCP terminal driver with UI semantics, not a generic bash execution skill.
Common Questions / FAQ
Who is Terminal MCP Server for?
Solo and small-team developers building TUIs or interactive CLIs who want Claude Code or Cursor to drive and test those apps like Playwright does for the web.
When should I use Terminal MCP Server?
Use it in Ship when you need repeatable agent tests for terminal apps, CI headless runs, or structured control of menus and prompts during debugging.
How do I add Terminal MCP Server to my agent?
Add the io.github.aybelatchane/mcp-server-terminal entry from the MCP registry to your Claude Code or Cursor MCP config, install the published package for your platform, and restart the client so the ten terminal tools load over stdio.