MCP server that gives AI agents real interactive terminal sessions
Fixes the core gap: AI agents can now actually use interactive CLIs, not just dump output.
Interactive terminals for AI agents, built for what you can't --yes away. SSH+MFA, GRUB/U-Boot, debconf installers, SOL/serial consoles, fsck, cryptsetup, pdb/gdb, apt, certbot and even Vim in tmux-backed sessions. Agent-driven, human-assisted for secrets/MFA. Single-file Python. Agent Skill. CI with 680+ tests. BSD License.
Agents finally handle SSH/pdb/vim—680 tests, tmux-backed, single-file Python.
AI agent developers, infrastructure automation engineers, DevOps teams
Expect (TCL) · pexpect (Python) · screen scripting
It supports in-band file transfer: the agent can move files through the terminal stream itself (same channel as the interactive session), which is useful when the agent doesn’t have scp/sftp, shared volumes, or direct filesystem access across boundaries.
Recent example: My Claude Opus was SSH'd into a server and ended up at a Firejail shell running inside a Docker container. It pushed a Python file in via term-cli, moving it across SSH → Docker → Firejail over the terminal channel, and explicitly commented how it was surprised this worked end-to-end.
And it comes with the companion tool term-assist, so agents can bring in their human to handle credentials and MFA: https://www.youtube.com/watch?v=A70tZEVqSOQ
Fixes the core gap: AI agents can now actually use interactive CLIs, not just dump output.
Headless PTY daemon lets AI agents control ncurses apps like k9s without a GUI.
Puppeteer for TUIs using libghostty-vt—finally agents can drive vim without screen scraping.
Headless xterm rendering lets agents navigate TUI apps like vim without stalling.
Terminal portfolios sound fun, but they're novelty—doesn't replace web portfolios.
Parallel SSH execution with AWS SSM connectors beats writing custom scripts.