xAI Ships Grok Build 1.0, an Agentic CLI That Operates Your Computer

Grok Build 1.0 is xAI's entry into the agentic-CLI category alongside tools like Claude Code and OpenAI's Codex. The command-line tool operates a user's computer—executing multi-step tasks, running commands, and orchestrating tools—following its May beta. The 1.0 release added dashboard summaries and integrated the newly released Grok 4.6 flagship, with a 1.0.5 patch on August 15.

The agentic-CLI mechanism gives the model direct access to the local environment, letting it act rather than just suggest. This is the same pattern driving the week's broader agent theme: from AWS AgentCore autonomous payments to Copilot's long-running agents, the frontier is shifting from chat to action.

Competitively, xAI is racing to match Anthropic's Claude Code (which is expanding weekly limits through August 31) and OpenAI's Codex harness, which Greg Brockman highlighted powering a tax-prep pilot that processed 7,000 returns and cut prep time by about a third. A capable local agent tied to Grok 4.6's low pricing could appeal to cost-sensitive developers.

The risks are the standard agentic-CLI ones: a model that operates your machine can cause real damage, and Grok 4.6's February knowledge cutoff limits its awareness of recent tooling. Security-conscious teams will want sandboxing and permission controls. Watch how Grok Build's reliability compares to Claude Code and Codex in real developer workflows.