Stakpak 3.0 CLI

We'd reach for Stakpak when we need an AI agent that can touch production infrastructure without risking credential leaks. The secret substitution design—where the LLM never actually sees your keys—isn't just a nice add-on; it's a fundamental safety feature that sets it apart from generic AI assistants. For teams that are already comfortable with the terminal and Docker, Stakpak provides a truly autonomous runtime that schedules tasks, runs health checks, and notifies via Slack/Telegram/Discord. Where it bites is the lack of a graphical interface. This is a CLI-only tool, so less technical team members or managers who prefer dashboards will struggle. Additionally, while the open-source nature is a huge plus for auditability and customization, it also means you're responsible for hosting and maintenance—no managed SLAs, no support team. If your infrastructure requires enterprise-grade support or you're not comfortable managing your own VMs, Stakpak isn't for you. Compared to a hosted PaaS like Heroku or Railway, Stakpak gives you more control and avoids lock-in, but you trade away convenience and operational overhead. Compared to generic AI coding assistants like GitHub Copilot, Stakpak is purpose-built for infrastructure tasks but lacks broad code generation capabilities. It's a niche tool, and it knows it. In practice, the setup is straightforward: install with curl, run stakpak init, then stakpak autopilot up. The preflight checks catch common issues before they cause downtime. The profile configuration lets you fine-tune model selection, auto-approval, and tool access per task, which is great for multi-environment setups. A real-world caveat: you need at least 2GB RAM and Docker, so it won't run on tiny instances. Also, the cron scheduling works well for periodic