OpenAI Codex CLI introduces the new `/goal` workflow command, allowing developers to set and monitor long-running, autonomous coding tasks directly inside their terminal.
OpenAI developer Dominik Kundel shared his experiences experimenting with the new `/goal` workflow command in the OpenAI Codex CLI. The `/goal` command (introduced in Codex CLI version 0.128.0) transitions the tool from a standard turn-based chat assistant to an autonomous agent capable of working persistently toward a defined, verifiable objective. Once set, the agent operates in a continuous loop of planning, acting, testing, and self-correcting until the objective is reached, with the goal state surviving terminal restarts.
The developer terminal is evolving from a command prompt into an orchestration center for autonomous agents, and OpenAI's persistent `/goal` command confirms that turn-based coding assistants are rapidly becoming obsolete.
* Turn-based to Agentic: By executing a continuous plan-act-test-verify cycle, `/goal` eliminates the need for constant "human-in-the-loop" prompting.
* Resilient Execution: The persistence of goal state across CLI restarts and model swaps makes it viable for complex, hours-long migrations and refactors.
* Converging Workflows: This feature aligns the Codex CLI with similar autonomous capabilities in tools like Claude Code and Cursor, highlighting the industry's standardization of goal-driven terminal interfaces.
DISCOVERED
2h ago
2026-06-04
PUBLISHED
2h ago
2026-06-04
RELEVANCE
AUTHOR
OpenAIDevs