Launching today
Grok Build
Agentic CLI for coding, building, and workflow automation
4 followers
Agentic CLI for coding, building, and workflow automation
4 followers
Grok Build is an agentic CLI that plans, delegates to parallel subagents, and executes coding and automation tasks from the terminal. Currently in early beta for SuperGrok Heavy subscribers.
xAI just shipped a terminal-native coding agent that runs parallel subagents, and it's one of the more structurally interesting CLI tools in this space right now.
What it is: Grok Build is an agentic CLI from xAI for professional software engineering, app building, and workflow automation that runs directly from your terminal.
Most coding agents today operate as IDE sidecars or chat wrappers. The limitation is sequential execution and single-context operation. Grok Build takes a different architectural approach: for complex tasks, it breaks work into a plan, lets you approve or rewrite that plan step by step, then delegates execution to parallel subagents that can work in separate git worktrees simultaneously. Every change surfaces as a clean diff.
What makes it different: The parallelism is structural, not cosmetic. Subagents can split research, build, and review work across tasks concurrently rather than linearly. It also picks up your existing repo conventions, AGENTS.md, plugins, hooks, skills, and MCP servers out of the box. Headless mode (-p flag) lets you run agents inside scripts and automations, and full ACP support means you can build your own bot and agent orchestration apps on top of it.
Key features:
Plan mode with step-by-step approval and full rewrite option before execution
Parallel subagents with deep worktree integration for large tasks
Headless mode for scripting and pipeline automation
MCP server support and plugin marketplace
Full ACP support for building custom bots and agent orchestration layers
In-CLI feedback via /feedback command
Benefits:
Complex multi-file tasks broken into reviewable, approvable steps
Parallel execution reduces time on large codebase operations
Integrates with existing project conventions without reconfiguration
Extensible via plugins and skills across a team
Who it's for: Professional software engineers and technical operators who run complex, multi-file coding tasks and want an agent that executes from the terminal with approval checkpoints rather than a chat interface.
Worth noting: this is early beta and currently gated to SuperGrok Heavy subscribers ($299/month, introductory at $99/month for six months). That's a narrow slice of the PH audience today. But the architecture here is worth watching as the access model almost certainly broadens.