iterm-mcp

iterm-mcp turns Claude into a proper terminal operator for iTerm2 on macOS. It executes shell commands, reads output, and sends control characters - basically giving Claude the ability to drive your terminal like you would. I've used it to watch build logs in real time while debugging, and it's genuinely useful when you want Claude to monitor a long-running process or interact with a shell session that needs back-and-forth. The killer feature is that it's not just fire-and-forget. Claude can send control sequences (Ctrl-C, Ctrl-D) and read what comes back, which means it can handle interactive workflows. I've had it SSH into a remote box, tail logs, grep for errors, and report back without me touching the keyboard. That's the kind of thing that saves you from context-switching when you're deep in something else. Quirks: it's macOS-only and iTerm2-specific. If you're on Linux or using a different terminal, this won't help. The setup is straightforward - npx gets you running fast - but you need to understand that Claude is literally controlling a terminal session. That means it can run destructive commands if you're not careful with your prompts. I wouldn't give it carte blanche on a production machine. It's also worth noting that this is most useful when you're already in Claude Desktop or a supported host and you want to offload repetitive terminal work. If you just need to run a single command and get the output, you're probably better off pasting it yourself. But for watching build processes, monitoring logs, or automating SSH workflows where you'd otherwise be tabbing back and forth, it's a genuine time-saver. Skip this if you're not on macOS, don't use iTerm2, or if your terminal workflows are simple enough that copying and pasting is faster. It's for people who spend a lot of time in the terminal and want Claude to handle the boring bits while they focus on the actual problem.