Codex CLI 0.128.0 adds /goal

OpenAI has released version 0.128.0 of its Codex CLI coding agent with a significant new feature that enables autonomous goal-driven development workflows. The /goal command allows developers to set high-level objectives, after which the AI agent automatically loops through iterative processes until either completing the task or exhausting its configured token budget. This development represents a meaningful step toward more independent AI-assisted coding capabilities.

The latest Codex CLI update introduces an implementation similar to the Ralph loop architecture, enabling extended autonomous execution. When developers specify a /goal, the system enters a continuous iteration cycle where Codex evaluates progress against the stated objective. The agent performs multiple rounds of reasoning, code generation, and assessment before concluding when it determines the goal is achieved or when token constraints are reached. This approach contrasts with previous versions that typically required more direct user intervention between each development step.

• Reduced Developer Friction: Developers can now specify objectives at a higher abstraction level, reducing the need for step-by-step guidance and prompting throughout development cycles

• Token Economy Trade-offs: The token budget limitation creates practical constraints on task complexity, requiring developers to balance ambition with resource allocation

• Autonomous Workflow Adoption: The feature accelerates adoption of autonomous AI agents in development environments, potentially reshaping how teams structure coding workflows

• Iterative Improvement Patterns: The looping mechanism enables agents to refine solutions across multiple iterations, potentially improving code quality and completeness compared to single-pass generation

• Competitive Pressure: This release positions OpenAI's Codex CLI against other autonomous coding tools, escalating the race toward more independent AI development assistants

The introduction of goal-based autonomous execution in Codex CLI represents a notable maturation of AI coding assistants. By enabling tasks to persist across multiple iterations with minimal human intervention, OpenAI demonstrates confidence in their agent's ability to self-evaluate and iteratively improve solutions. As these capabilities develop, developers and teams must navigate balancing increased automation with oversight and quality assurance, ensuring that autonomous goal completion aligns with project requirements and security standards. This feature will likely influence how organizations integrate AI coding tools into their development pipelines.