Grok Build gains /goal, an autonomous mode for long tasks

xAI launches /goal for Grok Build, an autonomous execution mode using Composer 2.5 and Grok Build 0.1 to plan, implement, and verify long coding tasks.

xAI introduces /goal into Grok Build, its code agent, a long-running autonomous execution mode designed to handle larger implementation tasks. The feature is available now, with the tool installing via command line.

The principle is straightforward: you set a goal, and Grok Build takes over. It plans an approach, breaks down the work into a list of tasks, then executes, with the user able to continue providing instructions along the way. Where a typical coding session involves back-and-forth between execution and verification, /goal continues until the task is completed and verified, whether it's reviewing code, inspecting web pages, or launching scripts. To do this, the mode orchestrates a series of agents dedicated to planning, implementation, and verification, with sub-agents working in parallel across multiple turns.

A few commands allow users to monitor and control a long-running task: a live progress panel, a pause that preserves the goal, a resume, or abandoning it. Once the goal is achieved, the panel switches to a completed state, with every box checked. Under the hood, /goal natively leverages two in-house models, Composer 2.5 and Grok Build 0.1.