SpaceXAI launches Grok Build terminal coding agent
The Rust-based harness supports interactive, headless, and embedded modes via the Agent Client Protocol, with binaries available for major operating systems under an Apache 2.0 licence.
SpaceXAI has released Grok Build, a terminal-based artificial intelligence coding agent and harness designed to streamline software development workflows. The tool features a fullscreen, mouse-interactive text user interface written in Rust, allowing developers to understand codebases, edit files, execute shell commands, search the web, and manage long-running tasks within a single environment.
The software is engineered to operate in multiple configurations, supporting interactive, headless, and embedded modes through the Agent Client Protocol. This flexibility enables use cases ranging from direct developer interaction to headless operation for scripting and continuous integration pipelines, as well as embedding within external editors. Prebuilt binaries are available for macOS, Linux, and Windows.
Hosted on GitHub under the xai-org organisation, the repository contains the Rust source code for the grok command-line interface and its agent runtime. The source is synced periodically from the SpaceXAI monorepo. While the binary artifact is named xai-grok-pager, official installations ship the tool as grok. Documentation is accessible at docs.x.ai/build/overview and is also included with the pager crate.
Upon first launch, the application opens the user’s web browser to authenticate. The project includes extensive features such as keyboard shortcuts, slash commands, configuration options, theming, support for MCP servers, skills, plugins, hooks, and sandboxing capabilities. The root Cargo.toml file is generated and intended to be treated as read-only, with developers advised to edit per-crate Cargo.toml files instead.
SpaceXAI has stated that external contributions to the project are not accepted. The first-party code in the repository is licensed under the Apache License, Version 2.0, while third-party and vendored code remains under its original licences. The release marks a significant step in providing open-source, terminal-native AI assistance for coding environments.
