Skip to main content
Back

xAI Opens Grok Build's Agentic Coding Tool to All Premium Users

xAI's terminal coding agent Grok Build opens its beta to all SuperGrok and X Premium+ subscribers, bringing Plan Mode, parallel subagents, Imagine, and CI-ready headless scripting.

Repo, Full Access
  • Beta now open: Grok Build is available to all SuperGrok and X Premium+ subscribers, not just the earlier limited access group.
  • Plan Mode: The agent shows a full, editable plan before touching any file -- every write is blocked until you approve.
  • Parallel subagents: Complex tasks are delegated to specialized subagents running simultaneously in isolated Git worktrees.
  • Zero-migration compatibility: AGENTS.md, CLAUDE.md, MCP servers, hooks, and skills from Claude Code all work out of the box.
  • Headless and CI-ready: The -p flag enables scripted, non-interactive runs for pipelines and automations.
  • API access: The underlying grok-build-0.1 model is also available directly via the xAI API for custom integrations.

xAI just widened the door on Grok Build, its terminal-based coding agent. After a limited early access period, the beta is now open to all SuperGrok and X Premium+ subscribers. That means the full feature set -- Plan Mode, image and video generation via Imagine, and the CLI for building automations and orchestrators -- is available to a much larger audience.

What Grok Build actually is

A chatbot responds to a single prompt with a single output. An agent, by contrast, is given a goal, plans multi-step actions to achieve it, executes those actions (file edits, code runs, tool calls), observes results, and replans when reality diverges from the plan. Grok Build is firmly in the second category. It runs directly in your terminal. You install it with a single command, point it at your project, and start issuing natural language instructions. Behind the scenes, it uses the Grok language model and a custom orchestration engine to plan, write, edit, and test code.

The release brings three major new capabilities: Plan Mode for structured task planning, Imagine for AI-powered image and video creation, and a CLI for building automations and orchestrators.

Plan first, edit later

The headline feature is Plan Mode, and it addresses the most common failure mode of agentic coding tools: the agent starts executing and makes a mess before you realize what it's doing. The plan viewer makes it easy to architect complex projects. In plan mode, every edit is blocked until you approve -- you can approve the plan, comment on individual steps, or rewrite it entirely.

Plan Mode is the feature xAI emphasizes most. Instead of a linear text plan, it generates a graph of sub-tasks with per-node state displayed in the TUI, and can activate parallel sub-agents on independent branches. You describe an objective, Grok builds a structured plan with files to touch and commands to run, and you approve before execution.

Parallel subagents and worktree isolation

For larger tasks, Grok Build delegates work to specialized subagents that run in parallel. Grok Build also supports deep worktree integrations, and you can launch subagents in their own worktrees. This means parallel agents work on isolated branches of your repo simultaneously, so their edits can't collide with each other or with your main branch.

The CLI shows which subagents are active and what they're working on. It's not always faster than sequential processing -- coordination has overhead -- but for large multi-file changes it can cut completion time significantly.

The extensibility stack

Grok Build ships with a layered extensibility model that will feel familiar if you've used Claude Code or Codex CLI:

  • Skills -- reusable slash commands you can create from any session using /skillify, or install from a marketplace
  • Hooks -- scripts that fire on agent lifecycle events like file edits, tool calls, or subagent start/stop
  • MCP servers -- connect to Linear, Sentry, Grafana, Postgres, and more via the Model Context Protocol
  • Plugins -- bundles of skills, agents, hooks, and MCP servers behind a single install command
  • AGENTS.md / CLAUDE.md -- both are recognized automatically; no migration needed

Grok Build recognizes both AGENTS.md and CLAUDE.md. If you're migrating from Claude Code, your CLAUDE.md works without changes. This zero-migration compatibility with Claude Code's configuration format is a deliberate design choice -- it lowers the switching cost for developers coming from Claude Code.

Headless mode and CI pipelines

Headless mode (-p) allows easily running agents inside scripts and automations. The CLI also provides full ACP (Agent Client Protocol) support to build your own bots and agent orchestration apps. This is the path for wiring Grok Build into CI/CD pipelines, scheduled jobs, or custom orchestration layers.

The underlying model, grok-build-0.1, is also available directly on the xAI API, so you can drop it into your own agent loop, IDE integration, or tooling without going through the CLI at all:

from xai_sdk import Client
from xai_sdk.chat import user
client = Client(api_key="xai-...")
chat = client.chat.create(model="grok-build-0.1")
chat.append(user("Refactor this function to handle null inputs."))
print(chat.sample().content)

Getting started

Grok Build is now available in Beta for all SuperGrok and X Premium+ users. Installation is a single command that works on macOS, Linux, and WSL:

curl -fsSL https://x.ai/cli/install.sh | bash

After installing, run grok inspect inside your repo to see what config, skills, plugins, hooks, and MCP servers the agent discovered before you trust it with a real change. As of now, Grok Build CLI requires a SuperGrok or X Premium Plus subscription. There's no public free tier or trial confirmed on xAI's site.

This is an early beta, and your feedback is the fastest way to make it better. Type /feedback in the CLI to send bugs, requests, and reactions straight to the team. The community has already started building a library of reusable skills and AGENTS.md templates on GitHub, which is a good starting point for teams that want to hit the ground running.