Back to stories
Tools

xAI Launches Grok Build, a Claude Code Rival With 2M-Token Context and 8 Concurrent Agents

Michael Ouroumis2 min read
xAI Launches Grok Build, a Claude Code Rival With 2M-Token Context and 8 Concurrent Agents

xAI launched Grok Build on May 15, 2026, its first dedicated coding agent and command-line interface, positioning the tool directly against Anthropic's Claude Code and other agentic developer assistants from OpenAI and Google. The company describes it as "a powerful new coding agent and CLI for professional software engineering and complex coding work," and is restricting early access to SuperGrok Heavy subscribers on its $300-per-month tier, with an introductory rate of $99 per month for the first six months.

The release is notable because Elon Musk had previously acknowledged that xAI was trailing Anthropic and OpenAI on coding capability, and Grok Build represents the company's most concrete attempt to close that gap.

What Grok Build does

Grok Build is an agentic CLI — it does not simply answer prompts but takes actions inside a developer's environment. According to xAI, the tool can plan projects, write and edit files, execute shell commands, and build complete applications from natural-language instructions. It is available at buildbygrok.com and integrates with Visual Studio Code for developers who prefer a GUI-adjacent workflow.

A distinguishing feature is a "plan mode" that lets developers review, edit, and approve a step-by-step plan before the agent touches the codebase — a guardrail aimed at reducing surprise edits in larger repositories.

Under the hood

Grok Build runs on a Grok 4.3 beta model with what xAI calls a 16-agent Heavy architecture. In practice, the system can spawn up to eight concurrent agents that simultaneously plan, search documentation, and write code. The agent ships with a 2 million token context window, which xAI says is enough to hold sizable codebases in working memory without aggressive retrieval.

Those specs put Grok Build in the same conversation as Claude Code on context length and parallelism, though independent benchmarks comparing real-world coding performance had not yet surfaced at launch.

Strategic context

The coding-agent race has intensified rapidly. Anthropic, OpenAI, Google, Cursor, and Cognition have all pushed major agentic developer products in recent months, and the market is increasingly bifurcating between IDE-integrated assistants and terminal-native agents. Grok Build firmly chooses the latter camp, leaning into a local-first, CLI-first workflow.

The launch also lands in a turbulent period for xAI. The company was folded into SpaceX in a February 2026 share-swap merger and has reportedly lost more than 50 researchers and engineers since, raising questions about its execution bandwidth even as it ships new products.

Implications

For enterprise developer tooling buyers, Grok Build adds a fourth credible agentic CLI to evaluate alongside Claude Code, OpenAI's Codex CLI, and Google's coding agents — but the $300-per-month entry point will limit early adoption to teams already paying for SuperGrok Heavy. For xAI, the bigger question is whether a long context window and concurrent agents translate into the kind of repo-level reliability that has made Claude Code the de facto benchmark. Early-beta feedback from SuperGrok Heavy subscribers over the next several weeks will be the first real test.

Learn AI for Free — FreeAcademy.ai

Take "Prompt Engineering Practice" — a free course with certificate to master the skills behind this story.

More in Tools

OpenAI Brings Codex to the ChatGPT Mobile App on iOS and Android
Tools

OpenAI Brings Codex to the ChatGPT Mobile App on iOS and Android

OpenAI is rolling out its Codex coding agent inside the ChatGPT mobile app, letting developers review, approve, and dispatch coding tasks from iPhone and Android while the work runs on a desktop machine.

13 hours ago3 min read
GitHub Launches Standalone Copilot Desktop App for Agentic Development
Tools

GitHub Launches Standalone Copilot Desktop App for Agentic Development

GitHub opened a technical preview of its native Copilot desktop app on May 14, giving developers a dedicated workspace for running multiple coding agents in parallel — each in its own isolated git work tree.

15 hours ago3 min read
Microsoft's MDASH AI Hunts Windows Bugs — Tops CyberGym Benchmark and Catches 16 Flaws Humans Missed
Tools

Microsoft's MDASH AI Hunts Windows Bugs — Tops CyberGym Benchmark and Catches 16 Flaws Humans Missed

Microsoft unveiled MDASH, a multi-model agentic AI security system that orchestrates more than 100 specialized agents to discover Windows vulnerabilities — finding 16 flaws in this month's Patch Tuesday and topping the CyberGym benchmark at 88.45%.

19 hours ago2 min read