How to write AI agent loops in Claude Code and Codex

2026-06-17 · Source: How I AI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, extended

Summary

The article introduces AI agent loops in Claude Code and Codex as an advanced method for autonomous AI prompting, moving beyond traditional human-triggered interactions. It details four types of automated prompting: heartbeat (scheduled tasks, e.g., every 5 minutes), cron (time-specific schedules), hooks (event-triggered, like email receipt), and goals (outcome-driven loops that run until validation). Effective loop implementation requires foundational elements such as work trees for isolation, skills for common tasks, plugins/connectors for tools, sub-agents for federated work and validation, and state tracking. Practical use cases include daily aging PR reviews in Claude Code and skill deepening automation in Codex. A key warning highlights that loops can be expensive due to token consumption, emphasizing the need for precise prompt and goal definition to ensure efficiency and cost control.

Key takeaway

For AI Engineers designing autonomous workflows, transitioning from manual prompting to automated agent loops in Claude Code or Codex is crucial. You should leverage scheduled routines or goal-driven loops to automate tasks like PR reviews or skill validation. Be meticulous in defining your loop prompts and validation criteria to prevent excessive token consumption and ensure your agents achieve desired outcomes efficiently.

Key insights

AI agent loops enable autonomous, scheduled, or goal-driven prompting, extending beyond human-triggered interactions.

Principles

• Loops automate agent prompting via schedules or goals.
• Effective loops require isolation, tools, and sub-agents.
• Precise goal definition is critical for cost and output quality.

Method

Design loops by defining a "job" for the agent, specifying triggers (heartbeat, cron, hook, goal), and equipping it with work trees, skills, plugins, sub-agents, and state tracking.

In practice

• Implement daily PR review routines in Claude Code.
• Automate skill validation using Codex sub-agents.
• Use scheduled tasks for morning briefings or inbox management.

Topics

• AI Agents
• Automated Prompting
• Claude Code
• Codex
• Workflow Automation
• LLM Cost Management
• Software Engineering

Best for: AI Architect, AI Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• How to design AI agent loops: schedules, goals, and subagents in Claude Code and Codex — AI agent loops enable autonomous, self-prompting workflows, extending automation beyond human-triggered inputs for continuous task execution.
• Stop Prompting. Start Looping. — AI agents can operate autonomously in loops, self-prompting to complete defined tasks and validate against goals.
• Claude Can Make Hundreds of AI Agents Now! 🤯 — Claude Code's workflow automates complex multi-agent development, planning, and rigorous QA testing.
• Claude vs. Codex isn't about code. It's about whether you steer or dispatch. — Claude Code and Codex exemplify distinct agent management paradigms: steering versus dispatching.
• I Watched the Entire Anthropic Workshop and Here Is a Recap — Engineers often underutilize agentic coding tools, treating them as autocomplete instead of autonomous execution engines.
• 7 INSANE loops you need to try right now — AI loops automate software development tasks by autonomously working towards defined, verifiable or LLM-judged goals.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by How I AI.