Introducing Grok Build: A Terminal-Based AI Coding Agent for Full-Screen TUI and Headless Automation

Grok Build arrived on GitHub as SpaceXAI's new terminal‑based AI coding agent. It ships as a full‑screen, mouse‑interactive TUI that can edit code, run shell commands, search the web, and manage long‑running tasks. The project also supports headless operation for CI pipelines and can be embedded in editors via the Agent Client Protocol (ACP). Built in Rust and released under Apache‑2.0, it targets macOS, Linux, and best‑effort Windows binaries. For developers who want AI assistance directly inside their terminal, Grok Build promises a tightly integrated workflow.

What happened

The Grok Build repository (github.com/xai-org/grok-build) contains the Rust source for the grok CLI/TUI and its agent runtime. The binary, named xai-grok-pager and distributed as "grok", runs as a fullscreen TUI that understands the current codebase, can edit files, execute shell commands, perform web searches, and orchestrate long‑running tasks. Users can also invoke the agent in headless mode for scripting or CI, or connect through the Agent Client Protocol (ACP) to embed it in editors and other tools.

Pre‑built binaries are provided for macOS and Linux, with Windows builds marked as best‑effort and not actively tested. The project pins the Rust toolchain via rust-toolchain.toml, and includes a protoc launcher for protobuf code generation. On first launch the tool opens a browser for authentication, while environments without a browser can use an API key set in ~/.grok/config.toml.

The agent ships with the grok‑4.5 model—the same model available via the xAI API—and supports custom models through the user‑level config file. Documentation covers keyboard shortcuts, slash commands, plugins, hooks, theming, and sandboxing, and the repository layout is read‑only for the workspace root.

Why it matters

By moving AI‑assisted coding into the terminal, Grok Build reduces context switches for developers who spend most of their day in a shell or Vim/Neovim. The headless mode enables automated refactoring, code generation, and testing within CI pipelines, potentially accelerating development cycles. Integration via ACP also opens the door for IDE plugins and custom tooling, extending the reach of the underlying grok‑4.5 model beyond the TUI.

How to think about it

Start by installing the pre‑built binary for your platform and run grok to complete the browser‑based authentication. For scriptable use, generate an API key and place it in ~/.grok/config.toml, then invoke commands like grok inspect to let the agent map your project. Choose the appropriate model with /model <name> inside the TUI or via the --model flag in headless mode. When integrating into an editor, use the Agent Client Protocol to forward requests, keeping the agent process separate for stability. Treat the TUI as a powerful REPL for code edits: experiment, accept or reject changes, and let Grok handle repetitive tasks such as dependency updates or boilerplate generation.

FAQ