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 stdioexposes 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 inspectshows what Grok discovered in the current directory. - Sandboxing & Permissions: Sandbox profiles,
--allow/--denypermission 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 -pin pipelines and bots.
Problems It Solves
- One agent, many surfaces: The same session works in the TUI, headless scripts, and editors via ACP.
- Long-running task management: Session resume, fork, and worktree support keep multi-hour agent runs manageable.
- 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:
- Rust-native performance: A fast single binary versus Node-based agents.
- First-party Grok integration: Deep integration with grok-4.5 and the xAI API.
- ACP-native: Built-in Agent Client Protocol support for editor workflows.
What Makes It Stand Out:
- The
grok wrapcommand that runs commands in a PTY forwarding clipboard writes via OSC 52. grok importto migrate sessions from Claude Code.- An Agent Dashboard for monitoring your agent runs.
Getting Started
Quick Start Guide
- Install: Download the prebuilt binary for your platform, or build from source (
cargo run -p xai-grok-pager-bin). - Authenticate: First launch opens a browser for authentication, or set
XAI_API_KEY. - Run:
grokfor the interactive TUI, orgrok -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
- Use worktrees for experiments:
grok --worktreeisolates risky changes. - Script with headless mode:
grok -pwith streaming JSON output fits CI loops well. - Scope permissions: Start with
--denyrules 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!
Related Tools
Reasonix
reasonix.io
DeepSeek-native AI coding agent for the terminal. A config- and plugin-driven harness in a single static Go binary, tuned around DeepSeek's prefix cache to keep token costs low across long sessions. 32,000+ GitHub stars.
Kimi Code CLI
www.kimi.com/code
Kimi Code CLI is Moonshot AI's open-source AI coding agent that runs in your terminal — reading and editing code, running shell commands, and planning its next steps with Kimi K2 models.
FastClaw
fastclaw.ai
Lightweight AI Agent runtime and Multi-Agent Framework written in Go. An Agent Factory that creates, manages, and runs AI agents with personalities (SOUL.md), memory, skills, and tools — single binary, any LLM, cloud-ready.