xAI Launches Grok Build: A Coding Agent to Rival Claude Code
xAI has launched Grok Build, a terminal-native agentic coding CLI powered by Grok 4.3 with a 2M token context window — xAI's first direct bid to compete with Claude Code and OpenAI Codex.
xAI
xAI Enters the Coding Agent Race With Grok Build Beta
xAI launched Grok Build today — a terminal-native agentic coding CLI that positions the company as a direct competitor to Anthropic's Claude Code and OpenAI's Codex CLI. The tool is in early beta and currently available only to SuperGrok Heavy subscribers paying $300 per month for the service.
xAI describes Grok Build as "a powerful new coding agent and CLI for professional software engineering and complex coding work." The company is soliciting feedback from early beta users to improve both the model and the product, and the launch itself represents the company's first serious move into the developer coding agent market — a space that Anthropic and OpenAI have been building out aggressively over the past year.
The release was personally promoted by Elon Musk, who posted calls for public beta testers on X on May 14, generating nearly 1.5 million views in minutes. SuperGrok Heavy users can install the beta with a single command: curl -fsSL https://x.ai/cli/install.sh | bash.
How Grok Build Works
Grok Build is powered by Grok 4.3 beta and uses a 16-agent Heavy architecture with a 2 million token context window — the largest among Western closed models and large enough to hold an entire large codebase in memory.
The key interaction model is plan-first execution. For complex tasks, Grok Build starts in plan mode: it writes a structured step-by-step plan, file by file, and waits for user approval before executing anything. Developers can approve the plan, comment on individual steps, or rewrite it entirely before any code runs. All changes surface as clean diffs.
For larger projects, Grok Build can spawn up to 8 concurrent parallel subagents that simultaneously plan, search documentation, and write code. It also supports deep worktree integrations, allowing subagents to run in their own isolated branches. Other capabilities include:
- Full MCP server support and a plugin ecosystem
- AGENTS.md, hooks, and skills compatibility — the same conventions used by Claude Code and Codex CLI
- VS Code integration for developers who prefer a GUI-adjacent workflow
- ACP (Agent Client Protocol) support for building custom orchestration layers on top of Grok Build
- Headless mode (
-pflag) for running agents inside scripts and CI pipelines
Context: xAI Playing Catch-Up on Coding
The launch is an explicit acknowledgement that xAI has been behind in the coding agent space. Elon Musk previously admitted the company had fallen short of competitors on coding performance, and earlier this year said he was rebuilding xAI "from the foundations up" after several co-founders departed. An executive reportedly told staffers to focus specifically on matching Claude's performance across coding tasks.
xAI was acquired by SpaceX in February 2026, forming SpaceXAI, but the merger has come with significant talent attrition. According to The Information, more than 50 researchers and engineers have exited since the merger, including key personnel in coding and AI training. Grok Build arrives as xAI tries to establish credibility in professional developer tooling while navigating that internal disruption.
Also taking effect today: xAI is retiring several older model slugs from the API. After May 15 at 12:00 PM PT, requests to retired endpoints — including grok-4-1-fast-reasoning, grok-4-fast-reasoning, grok-4-0709, and grok-code-fast-1 — will automatically redirect to Grok 4.3. Developers relying on these slugs should note they will be billed at Grok 4.3 pricing: $1.25 per 1M input tokens and $2.50 per 1M output tokens.
What's Still Unknown
Access to Grok Build remains gated behind the $300/month SuperGrok Heavy tier, making it one of the most expensive entry points in the AI coding agent market. xAI is currently offering an introductory deal at $99/month for the first six months, but has not announced a general availability timeline for lower subscription tiers.
The performance of Grok Build against Claude Code, Cursor, and Codex CLI in real-world engineering tasks has not been independently benchmarked. xAI is explicitly treating this as an early beta, and the quality of Grok 4.3 in agentic coding workflows will need validation before teams make infrastructure decisions around it.
