botctl does something obvious in hindsight: treat AI agents like any other long-running process. Created by developer Bret Logan, the open-source tool gives you a terminal dashboard and web UI to manage persistent bots that run Anthropic's Claude on a configurable loop. You define bots in BOT.md files, which combine YAML frontmatter for settings with markdown for system prompts. Start a bot, it spawns Claude with your prompt and tools, runs until it finishes, logs what happened, then sleeps until the next cycle. continuous agent loop. The practical features matter here. Session memory lets you resume a bot where it left off or send it a message mid-run to change direction. Say you've got a bot monitoring a GitHub repo for new issues and it starts going down a rabbit hole. You can message it to refocus without killing the run. Hot reload means editing your BOT.md takes effect on the next run with no restart. Skills are reusable modules you can find and install from GitHub to inject new capabilities into any bot. The tool tracks run counts and costs, which anyone running agents in production needs to watch. The web dashboard mirrors the terminal UI: readable terminal. Agent orchestration is getting crowded, and community feedback on the project notes the proliferation of similar tools. Some users have pointed out that botctl's landing page could communicate its value more clearly. This is a single-developer project treating agent management as a straightforward process control problem, not an enterprise platform play. That simplicity is the point. It runs on macOS, Linux, and Windows across AMD64 and ARM64, with installation via shell script or PowerShell. If you're running Claude-powered bots and managing them with cron jobs and shell scripts, botctl is worth a look.