xAI on May 14 released Grok Build, an early-beta terminal coding agent aimed at professional software engineering, bringing the Musk-led lab into a market already crowded with tools from OpenAI, Anthropic, GitHub and Cursor. The CLI installs via a single curl command on macOS, Linux and WSL, with a PowerShell path for Windows, and supports both an interactive terminal UI and headless scripting for integration via the Agent Communication Protocol.
In its release notes, xAI listed a plan-and-review workflow, diff review, multi-file edits, code search, terminal and sandboxed execution, git integration, and background task handling. The agent connects to developer-standard artefacts such as MCP servers, hooks, skills files and AGENTS.md documents. For larger tasks, it can spawn parallel subagents that operate inside separate git worktrees, a design the company says lets engineers parallelize complex code changes. The accompanying documentation, last updated May 24, underscores that the product remains in early beta and that features may evolve.
Access, however, has been a moving target. On May 14, Bloomberg reported the beta launch, and by May 15, GIGAZINE, Engadget and CIO Dive specified that Grok Build was available only to SuperGrok Heavy subscribers, the $300-a-month top tier. By May 25, xAI’s own product and news pages listed availability for "SuperGrok and X Premium Plus subscribers." The change, along with a documentation refresh on May 24, suggests a broadening beyond the initial high-end band, though the company did not respond to questions seeking confirmation of the current eligibility rules.
Under the hood, Grok Build runs on a dedicated model, Grok Build 0.1, which xAI is also offering through its API in early access. The model carries a 256,000-token context window and is priced at $1.00 per million input tokens and $2.00 per million output tokens, with aliases such as grok-code-fast-1. The model's slug first appeared in xAI’s release notes on May 19, and the API availability signals that xAI expects developers to build custom workflows around the agent.
The coding-agent segment has become a focal point for frontier labs, with each major release adding near-identical plan-and-review workflows, subagent delegation, and IDE integration. GitHub Copilot’s agent debuted in VS Code in May 2025, OpenAI open-sourced Codex CLI in April 2025 before expanding it into a cloud agent, Anthropic offers Claude Code with a large context window, and Cursor ships background agents. Google’s Gemini Code Assist rounds out the field. xAI enters the race with significant resources and an installed base of X subscribers, but Grok Build remains unproven. The X Premium Plus subscriber base could give xAI a large distribution channel, though it is unclear how many developers will adopt the CLI over existing tools. No independent evaluations or enterprise deployments have been made public, and no benchmark results have been released.
Key details are still unconfirmed. The precise ceiling on parallel subagents has not been officially disclosed. Some early reports speculated a limit of eight parallel subagents, but xAI has not confirmed any figure. Features such as "Arena Mode," which appeared in some secondary reports, remain unannounced by the company. Without third-party performance data, comparisons with established tools are speculative. The absence of benchmark scores—a standard practice among rival coding agents—makes it difficult to gauge Grok Build’s accuracy or speed.
For engineering teams, Grok Build adds a well-funded option to a market that is rapidly coalescing around a small set of interfaces—plan, code, review—delivered via terminal or IDE. Until early adopters share results and independent benchmarks emerge, the agent’s day-to-day reliability will remain an open question, and most organizations are expected to wait for validation before committing to the platform.