CLI tool
RongleCat/grok-app avatar
RongleCat/grok-app

Grok App: A Desktop Workbench for the Grok Build CLI, With Remote IM and a Pet Companion

Project brief: Desktop workbench for Grok Build CLI, sessions, projects, media, automations (Tauri 2 unofficial).

1,308 stars165 forksTypeScriptMIT

At a glance

What is it?
Grok App wraps the local Grok Build CLI in a Tauri 2 desktop client, adding project isolation, streaming sessions, file editing, and a long list of remote messaging bridges. It is not an official xAI product, and its value depends entirely on your existing Grok Build setup.
Who is it for?
Adopt Grok App if you already run the Grok Build CLI and want a desktop interface with project isolation, session forking, and remote IM control. Skip it if you expect a bundled model backend or a polished official product; you must install and sign in to the CLI yourself, and the README warns that full agent capabilities depend on that.
Can I use it commercially?
Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository received new commits within the last day.
What is it written in?
Mainly TypeScript, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.

DEEP OPEN-SOURCE ANALYSIS

What Grok App Actually Solves

Grok App addresses a specific gap: the Grok Build CLI is a terminal program, and managing multiple agent sessions, projects, and remote access from a terminal is tedious. The README describes it as a desktop workbench that wraps `grok agent stdio` via the ACP protocol. It is for developers who already use Grok Build and want a graphical layer for session management, file editing, and monitoring. It is not for people who want a standalone AI assistant; the README states clearly that it does not bundle proprietary model backends. All reasoning and tool execution run through your installed `grok` CLI. That is the core trade-off: you get a rich UI, but you still need the CLI installed, signed in, and maintained separately.

The ACP Protocol and Permission Model

The mechanism is straightforward: Grok App communicates with the CLI over the ACP protocol, which is the same stdio interface used by `grok agent stdio`. The README highlights granular permission tiers: Ask, Allow Once, Allow for Session, and YOLO mode. This is a real design choice. You can set per-project defaults, which means you can trust a project once and then let the agent run unattended. The default is Interactive Ask, which is safe but can become annoying during long sessions. The YOLO mode is the opposite: no prompts, full autonomy. The README also mentions workspace sandbox support, but it does not detail how the sandbox is enforced. That lack of detail is a limitation; you should verify whether the sandbox is a filesystem jail or just a permission flag.

Multi-Project Workbench and Git Worktree Integration

The workbench is organized around isolated project workspaces. Each project gets its own folder trust system and workspace isolation. The README mentions a virtualized sidebar, which suggests performance is a concern for large numbers of sessions. One notable feature is Git Worktree integration: the app automatically discovers linked worktrees, letting you switch working directories within a session without leaving the UI. Session forking is another strong point; you can branch off from any assistant reply with aligned context history. You can also attach up to three reference sessions via `/attach-chat` or drag-and-drop. These features are concrete and useful for developers who juggle multiple branches or experiments. The Kanban board (Needs Input / Working / Done) gives a visual overview of agent status, which is more informative than a plain log.

File Editing, Diffs, and Media Preview

Grok App includes an embedded CodeMirror 6 editor with live two-way file synchronization. The README says it has instant disk reload after agent modifications, which is important because the agent edits files on disk while you watch. You can review changes with visual Git diff, and accept or reject changes per file or in batch. That is a practical workflow for reviewing agent output. The media preview covers images, video, audio, PDF, and Office documents, which is broader than most CLI wrappers. There is also an AI asset generation feature via the Imagine skill, but the README does not explain how the skill is invoked or whether it requires a separate subscription. The integrated side-browser with Design Mode is for local web development previews; again, the README gives no setup details, so expect to experiment.

Remote IM Bridges and the Mobile Web Mirror

The most ambitious feature is the omnichannel remote IM bridge. It connects your local agent to Feishu/Lark, Telegram, Discord, Slack, DingTalk, WeCom, WeChat personal, QQ, Matrix, LINE, and Weibo. That is 11 or more channels. The idea is that you can monitor, resume, and steer your agent from your phone via a messaging app. The README mentions slash commands like `/p` for project switch and `/r` for resume. There is also a token-gated mobile web mirror and a loopback REST API (GET /v1/sessions, POST /v1/sessions/{id}/turns) for scripting. This is a genuinely different approach from most CLI wrappers, which are local-only. However, the README does not describe how the bridges are configured or whether they require a public tunnel. It mentions Cloudflare Quick Tunnel as compatible, so you will need to set that up yourself.

Getting Started: Installation and Configuration

The README gives a quick start path, but the details are truncated. It says the first-run setup wizard can assist with CLI installation, and UI-only development can be run with `GROK_APP_ACP=mock`. That environment variable is a concrete starting point for testing the interface without a real CLI. For production use, you need the Grok Build CLI installed and signed in. The app supports custom provider relays like OpenRouter, DeepSeek, and AI98PRO, with presets. It also mentions a non-destructive shared mode that protects existing `~/.grok` configurations. That is a thoughtful touch: you can try Grok App without breaking your existing CLI setup. API keys are stored in the OS keychain, which is a security plus. The README lists 15 built-in UI languages, so localization is not an afterthought.

Limitations and Wrong-Use Cases

The biggest limitation is that Grok App is a wrapper, not a model provider. If you do not have a Grok Build CLI subscription or a compatible relay, the app is useless. The README warns that full agent capabilities require a signed-in CLI. Another limitation is the dependency on the ACP protocol; if xAI changes the protocol, the app may break until it is updated. The README also mentions that it is an unofficial product, so there is no guarantee of long-term support. The remote IM feature is powerful but also a security risk: exposing your local agent to messaging platforms means you need to trust the token gating and the tunnel. The README does not provide a security audit or threat model. For users who only need a terminal-based agent, Grok App adds complexity without clear benefit. For users who need a GUI, it is a trade-off between convenience and maintenance burden.

Alternatives and How They Differ

The most direct alternative is to use the Grok Build CLI directly in a terminal emulator. That approach gives you the same agent capabilities without the GUI overhead, and it avoids the risk of protocol mismatches. Another alternative is a generic ACP-compatible desktop client, if one exists, but the README does not name any. A third alternative is to build your own automation using the CLI's stdio interface and a script, which gives you full control but requires more effort. The key difference is that Grok App bundles a lot of features (file editing, media preview, IM bridges, pet companion) that a terminal or a simple script does not. The pet companion is a differentiator, but it is also a sign of the app's scope: it is a product with personality, not a minimal tool. If you want a minimal tool, the CLI is the better choice.

Maintenance, Upgrade Cost, and License

The project is under active development, with releases v0.2.26, v0.2.27, and v0.2.28 pushed on consecutive days in August 2026. That suggests a fast release cadence, which is good for bug fixes but also means you may need to update frequently to keep up with protocol changes. The license is MIT, which is permissive and allows commercial use, modification, and redistribution, with attribution. The README does not mention a formal upgrade path or migration tool, so you should expect to reinstall or overwrite the app on each release. The data paths and configuration are described in the README, but the truncated text does not include the exact paths. You will need to consult the full README or the repository's docs to find where sessions and settings are stored. Given the unofficial status, you should also check the issue tracker for known problems before adopting it.

Editorial conclusion

Adopt Grok App if you already run the Grok Build CLI and want a desktop interface with project isolation, session forking, and remote IM control. Skip it if you expect a bundled model backend or a polished official product; you must install and sign in to the CLI yourself, and the README warns that full agent capabilities depend on that. Before adopting, verify that your Grok Build CLI version matches the ACP protocol version Grok App expects, and test the mock mode (GROK_APP_ACP=mock) to evaluate the UI without touching your real configuration. Also confirm that the keychain storage and custom relay modes work with your OS and that the non-destructive shared mode protects your existing ~/.grok files.

Official sources

  1. Official documentation
  2. Official README
  3. Project repository
  4. Release notes
Community notes

Community notes