What happened
In May 2026, on May 14, xAI, Elon Musk's AI company, launched Grok Build, its first coding agent that runs straight from the terminal. According to Bloomberg, which reported the launch the same day, the goal is to compete with established tools such as Anthropic's Claude Code.
xAI describes Grok Build as "a powerful new coding agent and CLI for professional software engineering and complex coding work." It launched as an early beta: at first, only SuperGrok Heavy subscribers (about $300 a month) had access. Later in May, xAI expanded the beta to all SuperGrok and X Premium+ subscribers.
What Grok Build does
Based on the official announcement and xAI's documentation, the main features are:
- Plan mode: for complex tasks, the agent drafts a plan before touching the code. Developers can approve it, comment on steps or rewrite it; afterwards, every change shows up as a clean diff.
- Parallel subagents: for larger tasks, Grok Build delegates parts of the work to specialized subagents (such as "explore," "plan" and "general-purpose") that run at the same time, each with its own context window, and can work in separate Git worktrees.
- Works with what you already use: it reads AGENTS.md files and supports plugins, hooks, skills and MCP servers.
- Headless mode (
-p) to run the agent inside scripts and automations, plus ACP support for building your own bots and orchestrators.
Alongside the CLI, xAI made the grok-build-0.1 model available on its API, also in beta.
An important correction: some social media posts claimed Grok Build runs "up to 8 subagents in parallel." We did not find that number in xAI's official announcement or documentation, so we treat that detail as unconfirmed.
The context
The launch came at a tough moment for xAI. Elon Musk had admitted the company was behind competitors in coding and, months earlier, said he was rebuilding xAI "from the foundations up" after several co-founders left. In February 2026, xAI was folded into SpaceX.
Meanwhile, the coding-agent market was crowded: Claude Code, OpenAI Codex, Google Antigravity and open-source projects like OpenCode were all chasing the same developers. Grok Build entered with a similar pitch: the agent plans, edits, tests and iterates while the human reviews and approves.
What this means for development teams
For companies evaluating coding agents, Grok Build's arrival reinforces a few good practices:
- Standardize project instructions in an
AGENTS.md: they now apply to several agents, not just one. - Require plan mode and diff review before any production change.
- Control permissions: Grok Build lets you block tools and subagents (for example with
--disallowed-tools), which matters in automations. - Protect secrets: terminal agents can access whatever your user can. Use least-privilege credentials.
- In beta, don't trust blindly: compare results with other tools before standardizing.
Sources
- xAI, "Introducing Grok Build": https://x.ai/news/grok-build-cli
- Bloomberg (2026-05-14): https://www.bloomberg.com/news/articles/2026-05-14/musk-s-xai-unveils-first-coding-agent-in-bid-to-rival-anthropic
- Engadget (2026-05-15): https://www.engadget.com/2173482/xai-coding-agent-grok-build/
- Subagents documentation: https://docs.x.ai/build/features/subagents
Want to adopt coding agents safely in your team? KGA IT offers a free consultation: kga-it.com.