Running AI tools on remote servers usually means one of two bad options: give an LLM agent direct SSH access to your infrastructure, or install AI tooling on every server you manage. Promptctl takes a third path.
The command-line tool works by wrapping SSH connections so that locally-defined LLM prompts appear as regular executable commands on the remote host. You run promptctl ssh user@server, and suddenly your custom prompts like llm-analyze-config or askai are available in the remote shell as if they were installed programs. The actual LLM inference (the process of generating a response from the model) happens on your local machine. The remote server never touches an API key or runs a model.
This solves a real headache for anyone managing multiple servers. Instead of configuring AI tools across a fleet of machines, you define your prompts once locally and carry them everywhere through SSH. The remote host stays clean, your API credentials stay local, and you get consistent behavior across every server you connect to.
The security angle is the most compelling part. Giving LLM tools SSH access has obvious risks, and spreading API keys across servers creates a larger attack surface. Promptctl keeps the AI layer entirely on the client side, which is the right default for most use cases. For developers and sysadmins who want AI assistance while debugging production systems, this is a much tighter security model than the alternatives.