Open‑source tool botctl lets developers run autonomous AI agents as long‑lived processes. Configuration lives in a YAML front‑matter block while the prompt sits in a markdown body, enabling quick edits without redeploying. A single command spawns Claude with the supplied prompt, tools, and workspace, then cycles through runs, logging output and sleeping between intervals.

Each run saves the session, so Claude can resume where it left off or receive new messages via “botctl start … --message”. Adjusting parameters such as max_turns in the BOT.md file takes effect on the next iteration without restarting the process. The tool also supports reusable skills modules fetched from GitHub, injecting extra capabilities into any bot.

Botctl includes a TUI and an optional web UI that mirrors all commands—start, stop, message, and live log streaming—at http://localhost:4444 or a custom port. By treating AI agents as services rather than one‑off scripts, developers can integrate code reviewers, Slack notifiers, or other automation directly into CI pipelines, reducing manual oversight.

Because sessions persist on disk, teams can audit every decision Claude made during a review, replaying logs for compliance checks. The modular skill marketplace also encourages community contributions, turning simple markdown prompts into plug‑and‑play extensions. In practice, botctl transforms a raw language model into a maintainable microservice that can be monitored, versioned, and scaled like any other backend component.