Grok-in-Codex

Grok-in-Codex

Enables Codex to delegate code reviews, coding tasks, image/video generation, and background investigations to Grok via MCP tools.

Category
访问服务器

README

Grok in Codex

Use Grok from inside Codex for code reviews, delegated coding tasks, background investigations, image generation, and video generation.

Codex stays the orchestrator. This plugin exposes a small MCP server that calls the local Grok CLI through a shared companion runtime.

Using Claude Code instead? Use the sibling plugin: grok-in-claude.

What You Get

Codex tool Purpose
grok_setup Check Grok CLI and authentication
grok_rescue Delegate investigation or fixes; write-capable by default
grok_review Structured read-only review of a tree, branch, or PR
grok_adversarial_review Challenge design, tradeoffs, and assumptions
grok_image Generate or edit images into .grok-media/image/
grok_video Generate short videos into .grok-media/video/
grok_status Show active and recent Grok jobs
grok_result Read final stored output for a job
grok_cancel Cancel a background job
grok_transfer Build context-transfer guidance for Grok

Skills included:

  • grok-routing for deciding when to delegate
  • grok-cli-runtime for companion runtime rules
  • grok-prompting for strong rescue prompts
  • grok-brand-media for image/video recipes

Requirements

  • Node.js 18.18 or later
  • Grok CLI on your PATH
  • Grok authentication via grok login
  • GitHub CLI (gh) only for grok_review with pull requests

Typical Grok CLI locations:

~/.grok/bin/grok
~/.local/bin/grok

Install

From GitHub:

codex plugin marketplace add stdevMac/grok-in-codex
codex plugin add grok@grok-in-codex

Then start a new Codex thread so the plugin skills and MCP tools are loaded.

Install Locally

From this repository:

codex plugin marketplace add /path/to/grok-in-codex/.agents/plugins
codex plugin add grok@grok-in-codex

Then start a new Codex thread so the plugin skills and MCP tools are loaded.

Run setup:

node plugins/grok/scripts/grok-companion.mjs setup

Or ask Codex to call grok_setup.

Quick Start

Ask Codex naturally:

Ask Grok to review this branch against main.
Use Grok to investigate why the auth tests are flaky.
Start a background Grok rescue job for the retry redesign.
Generate a 16:9 launch banner with Grok.
Show Grok job status.

Direct MCP tool examples:

grok_review base=main focus="auth, data loss, and race conditions"
grok_rescue prompt="investigate why npm test is failing" background=true
grok_status
grok_result jobId="task-abc123"
grok_image aspect="16:9" prompt="Dark developer-tool launch banner for Grok in Codex"
grok_video image="./.grok-media/image/hero.png" duration="6" prompt="gentle camera push-in"

Usage Notes

Rescue

grok_rescue is write-capable by default. Use readOnly: true for diagnosis only, and worktree: true for risky edits.

Useful options:

  • background: true for long-running work
  • model: "fast" for quick passes
  • model: "deep", effort: "high" for harder problems
  • worktree: true to isolate edits
  • check: true to ask Grok to verify before returning
  • bestOfN: 3 to run parallel attempts and select the best
  • resume: true, resumeSession, or fresh for session control

Review

grok_review is read-only and returns structured findings when parsing succeeds.

Examples:

grok_review
grok_review base=main
grok_review pr=123 focus="security and data-loss risks"
grok_adversarial_review focus="challenge this caching design"

Media

Images and videos are copied into project-local output folders:

.grok-media/image/
.grok-media/video/

Examples:

grok_image aspect="1:1" prompt="Simple geometric app icon, no text"
grok_image edit="./assets/logo.png" prompt="Make it monochrome with tighter padding"
grok_video refs=["shot1.png","shot2.png"] aspect="16:9" duration="6" prompt="product launch cutdown"

Video resolution is limited by the Grok model tier, commonly 480p.

Job Control

Background jobs can run concurrently. Use explicit job ids when more than one job is active.

grok_status
grok_status jobId="task-abc123"
grok_result jobId="task-abc123"
grok_cancel jobId="task-abc123"

How It Works

Codex
  -> Grok MCP tool
    -> node plugins/grok/mcp/server.mjs
      -> node plugins/grok/scripts/grok-companion.mjs <command>
        -> grok -p ... --output-format json|streaming-json

Job metadata is stored under:

~/.grok/codex-plugin/state/<repo-slug-hash>/

Set CODEX_PLUGIN_DATA to override the plugin data directory when running in custom environments.

Security

  • grok_rescue defaults to full tool auto-approval through Grok CLI write mode.
  • Use readOnly: true when you only want analysis.
  • Use worktree: true for higher-risk changes.
  • Reviews are read-only.
  • Media commands write under .grok-media/ and Grok session folders.

Development

npm test
node plugins/grok/scripts/grok-companion.mjs setup
node plugins/grok/scripts/grok-companion.mjs review --base main
node plugins/grok/scripts/grok-companion.mjs task --worktree --check "fix the failing test"
node plugins/grok/mcp/server.mjs

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选