Developers run continuous coding agents in headless mode with human approval workflows

A developer reported on Hacker News running three coding agents continuously over three days using headless command-line modes (Claude -p flag, Codex exec, OpenCode run) with dedicated human approval tooling replacing traditional UI-based control channels. The practice demonstrates emerging patterns for production deployment of coding agents with human oversight boundaries.