xAI’s Grok Build adds autonomous coding with /goal mode
xAI has added a new mode called /goal inside Grok Build, its terminal-based coding agent. Instead of guiding each step yourself, you hand the agent a larger goal and let it run autonomously—planning, executing, and verifying until the work is complete. The feature bundles long-running tasks into a single command, turning multi-step software changes into hands-off automation.
Inside the hands-off workflow
/goal operates within Grok Build, xAI’s command-line interface for software engineering. After installing the CLI and signing in, users invoke /goal with a one-line objective, such as “Migrate the auth module to the new API.” The agent breaks the task into a visible checklist, executes each step, and continues even if you step away. Commands like status, pause, resume, and clear let you steer the live run without derailing progress.
Built-in verification closes the loop
Unlike agents that merely edit files, /goal keeps running until the work is both built and verified. Verification can take three forms: code review, webpage inspection, or script execution. This built-in testing helps confirm that changes actually work before the agent declares the goal complete, reducing the risk of silent failures.
Who can use it and how to start
Access requires a SuperGrok or X Premium Plus subscription. Installation is a single command—curl -fsSL https://x.ai/cli/install.sh | bash—followed by account sign-in. Once active, /goal is designed for handoffs: describe the objective, and the agent owns the multi-step execution loop.
Source: MarkTechPost. AI-assisted editorial synthesis — TechnoExpress.