Grok (x) CLI Review: Free Coding Agent, Claude Code Plugins & LM Studio

Published 21/07/2026

Description

xAI has released Grok CLI, a command-line AI coding agent for working directly with your local projects. In this video, I install and test the Grok CLI on Windows, explore its coding-agent features, and connect it to a local model through LM Studio. Contents: 0:00 - Introduction 0:11 - Description 1:41 - Install 2:45 - First Login 3:53 - First Run 6:15 - Plan Mode 8:51 - / Command 9:58 - Market place / plugins 10:54 - MCP 11:27 - Hooks 12:23 - Custom Model 14:27 - To the limit 15:08 - Conclusion Grok CLI currently offers access to Grok 4.5 and provides a free usage allowance for users getting started. However, the exact practical limits, reset window, and usage reporting were not entirely clear during my testing. I test Grok CLI against a C# SharePoint upload project and explore how it analyzes a repository, creates a plan, proposes code changes, and works with different approval modes. This tutorial is useful if you are using: Installing Grok CLI on Windows Logging in and authorizing Grok Build Testing Grok CLI against an existing C# project Free usage allowance and observed limits Normal, Plan, Auto, and Always Approved modes Reviewing plans before allowing changes Slash commands, models, effort levels, and sessions Grok CLI plugins and marketplaces Reusing compatible Claude Code plugins MCP server configuration and Grok’s MCP connector Hooks for session, prompt, tool-use, and stop events Connecting Grok CLI to LM Studio Using an OpenAI-compatible local LLM endpoint Switching between a local model and Grok 4.5 A practical comparison of Grok CLI, Claude Code, and GitHub Copilot I also discuss an important privacy consideration: early users reported that project files and Git history could be uploaded during use. xAI subsequently changed the behaviour, but you should always understand what an online coding agent can access before using it with proprietary repositories, customer code, credentials, or se