---
title: Grok Build
description: Run xAI's Grok Build in Glade with your own xAI account. What it needs, how to sign in, and what works.
sidebar:
  icon: "/agents/grok.svg"
---

Glade runs your installed Grok Build CLI (`grok`) with the account it's signed in to.

| Requirement     | Value                                          |
| --------------- | ---------------------------------------------- |
| Minimum version | 1.0.0                                          |
| Install         | [Grok Build](https://docs.x.ai/build/overview) |
| Sign in         | `grok login`                                   |

## Set up

1. **Install Grok Build**

    Follow [xAI's guide](https://docs.x.ai/build/overview), then check that `grok` runs in a fresh
    terminal.

2. **Sign in**

    Run `grok login`.

3. **Pick a model**

    Choose a Grok model in the model picker. Glade reads each model's effort levels from Grok.

## What works

- **Chats, tools, approvals and slash commands**, the same way as Cursor and OpenCode.
- **Forks** copy the chat up to the reply into a fresh session.
- **`/compact`** runs Grok's own compaction.
- **Usage** shows the weekly credit pool, read from `~/.grok/auth.json`.

## Good to know

- Grok reports no context usage, so there's no context meter. Run `/compact` when a chat gets long.
- Messages sent mid-turn wait for the turn to finish.
- Grok rotates its sign-in tokens, so Glade never renews them. When the usage card asks, run
  `grok login` again.
- Grok can't delete a session, so a deleted chat's history stays in Grok's own storage.

## Under the hood

Glade talks to Grok Build over the [Agent Client Protocol](https://agentclientprotocol.com), on the
adapter it shares with Cursor and OpenCode, and hands it Glade's rules through its own `--rules`
option.