Grok Build logo

Grok Build

Visit

xAI's terminal-based AI coding agent powered by grok-4.5. A full-screen TUI that understands your codebase, edits files, runs shell commands, searches the web, and manages long-running tasks — interactively, headlessly, or via ACP.

Share:

Grok Build is xAI's terminal-based AI coding agent, powered by the grok-4.5 model. It runs as a full-screen TUI that understands your codebase, edits files, executes shell commands, searches the web, and manages long-running tasks. You can use it interactively, headlessly in scripts or bots, or embedded in other apps through the Agent Client Protocol (ACP). The agent runtime is written in Rust.

Key Features

  • Full-Screen TUI: A terminal-first experience with scrollback, modals, and a keyboard-driven prompt — built for deep work in the terminal.
  • Headless Mode: grok -p "prompt" runs one-shot queries for scripting and CI, with streaming JSON output support.
  • ACP Integration: grok agent stdio exposes Grok Build as an Agent Client Protocol agent, so editors like Zed can drive it.
  • Powerful grok-4.5 Model: The same model behind Grok Build is available via the xAI API for your own agent loops.
  • Custom Models: Add any model to ~/.grok/config.toml; grok inspect shows what Grok discovered in the current directory.
  • Sandboxing & Permissions: Sandbox profiles, --allow/--deny permission rules, and --always-approve (YOLO) for trusted environments.
  • Extensible: MCP servers, skills, plugins, hooks, experimental cross-session memory, git worktrees, session resume and fork.
  • Computer Use: Desktop automation on macOS via the built-in Computer sub-agent (accessibility-driven interactions).

Use Cases

Who Should Use This Tool?

  • xAI / Grok Users: Developers already invested in the Grok ecosystem who want a coding agent on the same model.
  • Terminal Developers: Anyone who wants a fast Rust-native TUI agent without Electron overhead.
  • CI Automation: Teams scripting agent runs with grok -p in pipelines and bots.

Problems It Solves

  1. One agent, many surfaces: The same session works in the TUI, headless scripts, and editors via ACP.
  2. Long-running task management: Session resume, fork, and worktree support keep multi-hour agent runs manageable.
  3. Trust boundaries: Granular permissions and sandbox profiles scope what the agent may touch.

Pricing

Plan Price Features
Open Source $0 MIT-style open source; bring your own xAI API key or auth. Model usage billed by xAI.

Advantages & Unique Selling Points

Compared to Competitors:

  1. Rust-native performance: A fast single binary versus Node-based agents.
  2. First-party Grok integration: Deep integration with grok-4.5 and the xAI API.
  3. ACP-native: Built-in Agent Client Protocol support for editor workflows.

What Makes It Stand Out:

  • The grok wrap command that runs commands in a PTY forwarding clipboard writes via OSC 52.
  • grok import to migrate sessions from Claude Code.
  • An Agent Dashboard for monitoring your agent runs.

Getting Started

Quick Start Guide

  1. Install: Download the prebuilt binary for your platform, or build from source (cargo run -p xai-grok-pager-bin).
  2. Authenticate: First launch opens a browser for authentication, or set XAI_API_KEY.
  3. Run: grok for the interactive TUI, or grok -p "Explain this codebase" headless.

Integration

Integrates with:

  • xAI API and grok-4.5
  • Zed and other ACP-supporting editors
  • MCP servers, skills, plugins, and hooks

Frequently Asked Questions

Can I use other models with it?

Yes — Grok Build supports custom models via ~/.grok/config.toml.

Does it work in CI?

Yes — headless mode (grok -p) is designed for scripts, automations, and CI pipelines.

Is it open source?

Yes, the source is on GitHub (xai-org/grok-build), with prebuilt binaries for macOS, Linux, and Windows.

Alternatives

If Grok Build isn't the right fit, consider these alternatives:

  • Reasonix: DeepSeek-native terminal agent in a single Go binary.
  • Claude Code: Anthropic's terminal coding agent (Dev Tools category).
  • CrewAI: Python role-based multi-agent orchestration.

Tips & Best Practices

  1. Use worktrees for experiments: grok --worktree isolates risky changes.
  2. Script with headless mode: grok -p with streaming JSON output fits CI loops well.
  3. Scope permissions: Start with --deny rules before enabling --always-approve.

Conclusion

Grok Build brings xAI's grok-4.5 into your terminal as a fast, Rust-native coding agent with TUI, headless, and ACP surfaces. If you're in the Grok ecosystem or want a performant terminal agent with strong editor integration, it's worth a look.

Comments

No comments yet. Be the first to comment!