Musk's xAI Launches Grok Build, a 2M-Context Coding CLI Aimed Square at Claude Code

xAI on Friday opened an early beta of Grok Build, a new agentic command-line tool for software engineering that the Elon Musk-led company is positioning as its direct answer to Anthropic's Claude Code and OpenAI's Codex. According to Bloomberg, the launch represents xAI's first dedicated coding agent and a clear admission that text-only Grok was no longer enough to compete in the most lucrative frontier of consumer AI: developer tooling.

The tool is gated behind SuperGrok Heavy, the company's $300-per-month tier, and ships with hardware-aggressive defaults. Grok Build is powered by Grok 4.3 beta, which xAI describes as a 16-agent "Heavy" architecture with a 2 million token context window — enough to load entire mid-sized repositories into a single session. The CLI can spawn up to 8 concurrent subagents that plan, search documentation, and write code in parallel, and it supports git worktree integration so each subagent can operate in its own isolated checkout.

The interaction model is built around a planning loop that will feel familiar to anyone who has used Claude Code. Users can start a session in plan mode, review the proposed steps, comment on individual items, or rewrite the plan entirely before any code is touched. Once approved, every change is presented as a clean diff for human review. Engadget noted that the design choices closely mirror conventions popularized over the last year by Anthropic and OpenAI's coding agents — a sign that the agentic-CLI form factor is rapidly converging across vendors.

The timing is pointed. Just one day earlier, OpenAI and Anthropic both moved aggressively to win developer mindshare — OpenAI offered two months of free Codex access while Anthropic lifted Claude Code rate limits by 50% on the same day. xAI's response is to skip the freebies and bet on raw capability behind a premium paywall, wagering that SuperGrok Heavy subscribers will tolerate the price tag in exchange for the largest context window and most aggressive parallelism on the market. Whether developers agree will be visible in xAI's telemetry over the coming weeks, since the early beta routes a built-in /feedback command straight to the engineering team.