Night Orders Studio’s Grok Bot Chief of Staff describes the same hybrid SpaceXAI ships: hand coding to Cursor cloud agents and a Grok Build terminal, while a desk of agents (Build, Legal, Tester, App Store) checks every build and one human signs off. Lingxi Li’s Engineer Bot on the Grok Bot Marketplace packages the supervisor lane—board work on the repo you name, launch cloud agents, watch PRs ~every 30 minutes, only ask you to merge. This workflow is listed under Grok-Bot, engineering, PR-review, coding-agents, HITL.
MIT Go daemon + git hook reviews every commit with the CLIs you already pay for—Codex, Claude Code, Cursor, OpenCode, and first-class Grok Build (grok / grok-build). Findings stay in a ledger until closed; agent hooks push open review IDs back into the session for /roborev-fix before context goes cold. Reviews stay on your machine.
Common xAI stack pattern (confirmed by @grok replies and power users): keep Grok Bot as the always-on teammate that plans, researchs, and hands off; route multi-file coding to Grok Build CLI (`curl -fsSL https://x.ai/cli/install.sh | bash`—plan mode, parallel subagents, skills) or Cursor cloud agents. Bot’s cloud computer can still clone/run code, but dedicated local/repo coding stays cheaper and more controlled on Build.
Documented consumer flow: drop a contract, lease, or estate-planning PDF into Grok Bot; it flags hidden fees and tricky clauses and can help draft a dispute letter. Mario Nawfal’s Oct 17 demo amplifies the same pattern Elon highlighted for legal-doc review.
Ledger’s Agent Intent queue lets background agents (Grok Bot, Muse, Hermes, OpenClaw, Dots) prepare actions—e.g. Grok Bot watches an inbox and drafts an invoice payment—while you review asynchronously and physically confirm on a Ledger signer Secure Screen. Agents never hold signing keys; coding agents can use Agent Stack CLI/skills too.