Is Grok Build open source?

Yes. xAI released it under Apache 2.0 in July 2026 — the full Rust source for the grok CLI, its terminal UI, agent runtime, tools, MCP support and sandboxing.

The harness is free and open under Apache 2.0. The model behind it is not — you still connect it to a model via API, so you pay for tokens, not for the agent.

Grok Build vs Claude Code, which is better?

Grok Build is open and self-hostable with real-time web and X data. Claude Code is closed but ships a broader MCP tool ecosystem out of the box. It comes down to open-and-inspectable versus most-integrations.

grok build vs claude code: which cli coding agent wins?

Grok Build and Claude Code are both terminal coding agents: they read your codebase, edit files, run shell commands, and ask before doing damage. The real split is what you can see and connect. Grok Build is open source (Apache 2.0) and written in Rust — you can read it and self-host it. Claude Code is closed but ships the deeper tool ecosystem. Open vs. batteries-included.

tool	license	language	leans on	pick it if
Grok Build	Apache 2.0 (open)	Rust	real-time web / X data	you want to read, audit, or self-host the agent
Claude Code	closed	Node	broad MCP ecosystem	you want the most integrations working out of the box

what is grok build?

xAI open-sourced Grok Build on July 15, 2026 under Apache 2.0 — the actual Rust source for its grok CLI. The xai-org/grok-build repo has the terminal UI, the agent runtime, the tool implementations, workspace and version-control handling, MCP support, sandboxing, and telemetry. It’s a coding agent that reads a codebase, edits files, runs shell commands, searches the web, and manages long-running tasks. There’s an interactive TUI, a headless mode for CI, and editor integration through the Agent Client Protocol. Alongside the code drop, xAI reset usage limits, turned data retention off by default, and said it would delete coding data kept before the change.

how is it different from claude code?

Same shape, different bet. Both are file-aware, conversational, diff-based, and gated behind human approval — you won’t feel lost moving between them. Where they split:

Open vs closed. Grok Build is Apache 2.0. You can read every line, fork it, self-host it, and see exactly what the agent does with your code. Claude Code is closed source. If “inspectable” matters to you, that’s the whole ballgame.
Ecosystem vs freshness. Claude Code’s MCP ecosystem is broader today — GitHub, Jira, databases, Slack connect out of the box. Grok Build leans on xAI’s real-time data (live web, X) to check current docs or library versions mid-build without switching context.
Language. Grok Build is Rust — one fast binary that’s easy to audit. Claude Code runs on Node.

is grok build a real tool or a code dump?

It’s the real harness, not a teaser. The repo is the same scaffolding behind xAI’s own coding CLI — TUI, runtime, tools, MCP, sandboxing — not a stripped sample. One catch worth stating plainly: open source doesn’t mean free frontier model. You still point the harness at a model through an API, so the agent is free and the inference is not. What you get for nothing is the scaffolding and the freedom to run it your way.

which should you use?

Read, audit, or self-host the agent → Grok Build. Apache 2.0 is the reason it exists.
Most integrations working today, zero setup → Claude Code. The MCP ecosystem is deeper.
Current web / X data while coding → Grok Build’s real-time access.
Don’t care about internals, just want it to drive → either. They feel the same in the terminal.

The interesting part isn’t which agent is smarter — the models move every month and that ranking is stale by lunch. It’s that xAI open-sourced the harness while most competitors keep theirs locked. The agent scaffolding — the TUI, the tool layer, the approval loop — is quietly turning into a commodity. The moat was never the wrapper. It’s the model, and now everyone can read the wrapper.