XAi's Grok Build CLI: Ambitious Interface Challenged by Model Immaturity and Steep Pricing

XAi, Elon Musk’s artificial intelligence company, has unveiled Grok Build, a new CLI-based AI agent intended to compete with leading platforms like OpenAI’s Codex/GPT and Anthropic’s Claude. Available across Windows, Linux, and macOS (including WSL), Grok Build offers a sophisticated Terminal User Interface (TUI) with mouse interaction, supporting various modes such as build, plan, and always apropos. Its feature set includes comprehensive project planning capabilities, exemplified by generating detailed plan.md files that consider file structures and leverage web searches for code inspiration. The agent also boasts support for en skills, en MD, and MCPs, alongside a marketplace for plugins and the ability to import existing configurations from other AI agents, streamlining migration. Furthermore, Grok Build implements advanced functionalities like sub-agents for concurrent execution of tasks and a loop command for iterative goal achievement, positioning it as a potentially versatile tool for developers.

However, initial practical evaluations of Grok Build reveal a significant disparity between its robust interface and the underlying model’s performance, especially considering its premium pricing. The “Super Grok Heavy” subscription, recommended for CLI use, is priced at $300/month (currently discounted to $99/month during beta), making it one of the most expensive AI agent subscriptions, surpassing Claude Max and Codex. Despite this high cost, tests indicate that the model frequently generates incomplete code and hydration errors in web applications, struggling to establish proper connections with local development tools. Its code generation capabilities have been likened to “older models, such as early versions of GPT or Opus,” suggesting a notable lack of maturity. While the TUI and multi-session support are commendable, the current state of the intelligent model renders Grok Build challenging for practical, large-scale application development, with some comparisons even favoring less expensive “Chinese models” for responsiveness. This preliminary assessment suggests that while Grok Build’s architecture and interface show promise, its core AI model requires substantial improvement to justify its high price and compete effectively in the crowded AI agent market.