Back to Gap Map Product / UX / Orchestration & agents

Grok Build

xAI
open source / Overall score: 3.7

Grok Build is a terminal coding agent that edits files, runs shell commands and searches the web from a full-screen TUI, headless scripts, or editors over the Agent Client Protocol. It adds parallel subagents, plan mode, an opt-in OS-level sandbox, MCP, skills, plugins and hooks, and can run other providers' models beside its default Grok models. SpaceXAI develops it internally and publishes the source as a periodically synced mirror.

Read against xai-org/grok-build, whose README and LICENSE name SpaceXAI while the docs, installer and API sit on x.ai, so the entry stays under xAI. The npm package is named in the vendor's enterprise docs rather than the README.

Openness

5 medium confidence
5.0
license
Apache-2.0(OSI)
source
public(GitHub mirror of an internal monorepo)
core features withheld
no — the tree at 2bdd1d6 carries one root Apache-2.0 LICENSE and no ee/, enterprise/, commercial/ or proprietary/ path

Grok Build's first-party code is Apache-2.0, and the public repository holds the Rust source for the CLI, its agent runtime and its enterprise policy settings. What SpaceXAI sells is model access through a Grok sign-in or an xAI API key, not features held back from the code.

Adoption

3 medium confidence
3.0

Adoption is the trailing month of npm downloads of @xai-official/grok, the package the vendor's docs offer for installing Grok Build; the per-platform binary packages it pulls in are not added. Installs through the install script on x.ai publish no count and are not included, so the figure is a lower bound on use.

Capability

4 medium confidence
4.0

Grok Build covers the full terminal coding agent, with parallel subagents in separate worktrees, plan mode, an OS-level sandbox, ACP for editors and other providers' models. It sits a step below OpenHands, which runs agents in containers, on VMs or in cloud backends and publishes benchmark results, while Grok Build runs on the user's machine with no benchmark of its own.

Verified 2026-10-19, except some scores, still awaiting re-confirmation