xAI’s Grok Build Introduces ‘/goal’ for Autonomous Coding Tasks

xAI has launched a new feature called ‘/goal’ within Grok Build, its terminal-based coding agent. This addition is designed to facilitate long-running, autonomous task execution for software development. The ‘/goal’ mode allows developers to assign a larger objective to the agent, which then plans, executes, and verifies the multi-step work independently until completion.

Traditionally, coding sessions with AI agents often involve a continuous loop of prompting, execution, and manual verification for each step. The introduction of ‘/goal’ aims to streamline this process by empowering the agent to manage the entire workflow autonomously. This includes internal verification steps such as reviewing generated code, inspecting web pages for behavioral changes, or executing test scripts to confirm the outcome.

Key facts

Feature	Description
Product	Grok Build (xAI’s terminal coding agent)
New Mode	/goal
Core Function	Autonomous, long-running execution of multi-step coding tasks with built-in verification.
Verification	Code review, webpage inspection, script execution.

How ‘/goal’ Works

When a developer initiates a task using the ‘/goal’ command within Grok Build, they provide a single objective, such as “Migrate the auth module to the new API.” The agent then takes ownership of this multi-step work. It generates a detailed plan, breaking the objective into a progress checklist. Subsequently, Grok Build executes items on this checklist one by one.

A significant aspect of ‘/goal’ is its emphasis on verification. Unlike agents that might only produce code and report success, Grok Build with ‘/goal’ continues working until the task is not only completed but also verified. This verification can involve the agent reviewing its own code, checking web pages to confirm desired behavior, or running scripts to test the results. This integrated verification ensures that the output is functional and meets the specified requirements before the agent marks the task as complete.

Benefits for Developers

For developers, ‘/goal’ offers the potential to offload complex, mechanical, and time-consuming coding tasks that involve multiple files and steps. Examples include module migrations, refactoring efforts, or implementing new features that require broad changes across a codebase. By automating these processes, developers can focus on higher-level design, architecture, and more creative problem-solving.

The system also provides observability and control during autonomous runs. Users are not locked out while the agent is working and can provide additional instructions if needed. A visible progress checklist allows for monitoring the agent’s progress. Furthermore, commands like ‘/goal status’, ‘/goal pause’, ‘/goal resume’, and ‘/goal clear’ offer granular control over the ongoing task, allowing developers to check progress, halt work, continue from where they left off, or cancel the goal entirely.

Grok Build Integration

‘/goal’ is a mode within Grok Build, xAI’s command-line interface (CLI) for software engineering. Grok Build operates directly against a local codebase, reading files and executing commands. It already supports various agent features, including a plan mode that requires approval before edits, delegation of larger tasks to specialized subagents, and compatibility with AGENTS.md, plugins, hooks, skills, and MCP servers. Access to Grok Build and the new ‘/goal’ feature requires a SuperGrok or X Premium Plus subscription.

This development from xAI underscores a growing trend in AI-assisted software development towards more autonomous and self-verifying systems. By handling the ‘observe–plan–act’ loop with built-in verification, ‘/goal’ aims to enhance efficiency and reliability in complex coding workflows.

Source: MarkTechPost, https://www.marktechpost.com/2026/06/22/xai-launches-goal-in-grok-build-adding-long-running-autonomous-execution-with-built-in-verification-for-multi-step-coding-tasks/