ivrylobs/grok-cc-plugin — explained in plain English
Analysis updated 2026-05-18
Delegate a coding task to a Grok worker while Claude approves risky actions.
Let a worker pause and ask for missing information instead of guessing a secret.
Resume a worker session after it was interrupted, keeping its memory intact.
Adjust how much autonomy a worker has using different grip levels.
| ivrylobs/grok-cc-plugin | 0xradioac7iv/tempfs | 52191314/web-agent-proxy-sdk | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 4/5 | 3/5 | 4/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Node 20+, a logged-in grok CLI account, and Claude Code restarted after plugin install.
grok-cc-plugin is a Claude Code plugin that lets Claude hand off coding tasks to a separate AI model, Grok, while staying in control of what Grok is allowed to do. Instead of letting Grok run completely on its own, the plugin routes every file change and every risky command through a review step. Claude reviews and can approve or deny each risky action before it runs, so a worker never silently deletes files or runs a dangerous command without a person or Claude signing off first. The README describes two of its central behaviors with runnable proof rather than just claims. First, if Grok is asked to do something it does not have enough information for, such as filling in a real production secret key, it is designed to pause and ask a question instead of making up an answer. Second, every tool call a Grok worker wants to run, like creating a folder or running a shell command, is checked against a policy before it executes, and a denied action never actually happens. The project includes an offline test suite that can be run locally to verify these two behaviors, plus an optional live test that uses a real Grok account. Once installed, a person delegates a task to a Grok worker with a command like /grok:work, and a background process keeps track of the task so Claude gets notified when the worker finishes, asks a question, or needs a decision on a risky action. Other commands let a person check on running workers, approve or deny their pending requests, fetch a finished result, resume a worker that was interrupted, or stop one entirely. A separate setting called grip level controls how much a worker can do without asking first, ranging from a strict mode where every in-project file write is held for manual approval, to a looser mode where only a short list of destructive actions require sign-off. Installing it requires Node.js version 20 or later, the separate grok command line tool logged into a Grok account, and Claude Code itself. The plugin is installed as a Claude Code plugin marketplace entry and needs Claude Code restarted afterward so its commands load.
A Claude Code plugin that delegates coding tasks to Grok while requiring approval before risky actions run.
Mainly TypeScript. The stack also includes TypeScript, Node.js, Claude Code.
No license information is stated in the README.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.