Grok Build hits 1.0: xAI's terminal coding agent goes stable
xAI's answer to Claude Code locks in a stability promise, plan mode, and eight parallel subagents — and Cursor now hands its agents your Gmail and Drive.
Copy markdown
A version number that's really a promise
Reaching 1.0 means xAI now commits to API stability, published deprecation policies, and predictable release cycles — so you can pin Grok Build in CI without weekly breakage. Safety got a lift too: every bash script is fully visible before it runs, and Esc or a stop command cancels a task cleanly.
Plan mode, then eight subagents in parallel
Grok Build proposes a step-by-step plan and waits for your approval before touching files, then fans large jobs out to up to eight subagents, each in an isolated Git worktree so parallel edits never collide. It slots in directly against Claude Code, Codex CLI, Copilot CLI, and Gemini CLI.
Your existing agent config just works
AGENTS.md, MCP servers, plugins, hooks, and skills all load out of the box, and headless mode authenticates from a single environment variable for servers and CI runners — dropping Grok Build into an existing pipeline is close to zero setup. It also speaks the Agent Client Protocol for wiring into other editors.
What it runs on, and what it costs
Grok Build drives xAI's grok models: grok-4.5 offers a 500K-token context at $2/$6 per million input/output tokens, while a lighter grok-build model runs cheaper for headless use. It's bundled into SuperGrok subscriptions, reportedly from about $99/month on the intro tier.
Elsewhere: Cursor gives its agents your Google Workspace
Cursor's new Google Workspace plugins let its agents pull context from and act on Gmail, Drive, and Calendar without leaving the editor — drafting replies, updating docs, and managing events as part of a coding session.