xAI has launched Grok Build, a code agent that operates directly from the terminal. Unlike a conventional chatbot, this tool plans, writes, executes, and corrects code autonomously in cycles. It installs with a single command and directly competes with Anthropic's Claude Code and OpenAI's Codex CLI.
Autonomous code cycles with two million token context 🚀
Grok Build uses the Grok 4.3 beta model and offers two operating modes: Normal, with one agent and 128 thousand tokens of context, and Heavy, which deploys 16 agents working in parallel with a context window of 2 million tokens. The agent analyzes the project, generates a plan, writes the code, executes it, and corrects errors without manual intervention. It is available for Linux, macOS, and Windows.
The assistant that doesn't ask permission to break your code 🤖
While chatbots ask you if you want to continue, Grok Build directly executes, fails, and corrects in a loop. You will no longer have an excuse to blame the compiler: now an autonomous agent will do the dirty work and, incidentally, take credit when it works. All you have left is to watch it from the terminal while it decides which files to delete.