xAI launches Grok Build, an agentic coding agent built on a dedicated 256k-context model

xAI launched Grok Build in May 2026, entering the coding agent market with a purpose-trained model — grok-build-0.1 — built specifically for software engineering workflows. The release positions xAI as a direct competitor to Anthropic's Claude Code, GitHub Copilot, and Cursor, with a developer-first architecture centered on a command-line agent and API-first design.

What's new

Grok Build is described in xAI's documentation as "a powerful and extensible coding agent" accessible through three interfaces: an interactive terminal UI (TUI) for hands-on developer sessions, a headless scripting mode for automation and bot use, and integration via the Agent Client Protocol (ACP), which allows third-party applications to embed Grok Build without custom API wrappers.

The underlying model, grok-build-0.1, carries a 256,000-token context window and is priced at $1.00 per million input tokens and $2.00 per million output tokens through the xAI API. Key technical features include:

• Natural language task input: developers describe a task in plain English; the agent inspects the project structure, identifies the relevant files, and makes targeted changes
• ACP integration: third-party platforms can embed the agent without custom wrappers, enabling plug-and-play adoption across developer tooling
• Headless scripting mode: supports fully automated workflows, enabling Grok Build to run inside CI pipelines or bots without user interaction

Access at launch was limited to the SuperGrok Heavy tier. The grok-build-0.1 model is separately available through the xAI API for teams building on top of it directly. Shortly after launch, xAI confirmed Grok Build 0.1 integration into Notion AI alongside the separately released Grok 4.3 model.

Context

The coding agent market has become one of the most commercially active categories in applied AI through early 2026. Anthropic shipped Claude Code as a standalone product and expanded it with multi-agent orchestration and Workflows. Microsoft deepened GitHub Copilot's agentic capabilities. Several third-party tools — Cursor, Codeium, Devin — have established substantial developer adoption.

xAI's approach with Grok Build differs from labs that repurpose general-purpose models for coding: grok-build-0.1 is purpose-trained for software engineering tasks, following the same pattern OpenAI used with GPT-Rosalind for life sciences. The ACP support treats the agent as infrastructure to be embedded rather than a standalone product, lowering the integration threshold for enterprise tooling.

At $1.00/$2.00 per million tokens, grok-build-0.1 is priced well below Claude Opus 4.8 ($5/$25 per MTok) and the upper end of GPT-5.5, targeting high-volume automated coding workflows where cost-per-edit matters.

Why it matters

A dedicated coding model from xAI signals that the company is moving beyond consumer Grok products toward developer tooling as a primary revenue channel. The purpose-trained approach — rather than a system-prompted general model — suggests differentiated performance on software engineering benchmarks.

The ACP design is worth attention: by standardizing the integration protocol, xAI is building an ecosystem layer. If Grok Build proves reliable in production, the ACP pathway makes it straightforward for other platforms to adopt it — reducing switching costs and increasing distribution beyond xAI's own surfaces.

For teams running automated coding workflows at scale, the $1/$2 pricing against a 256k context window is a meaningful competitive offer. The practical question is whether purpose-trained performance advantages over a system-prompted Opus 4.8 or GPT-5.5 are large enough to justify a platform migration — a judgment that will depend on per-task benchmark results teams run themselves.