mrkeyoor.com_
Fri 25 Sept 15:55 UTC
AI Toolsevaluationupdated 25 Sept 2026

desktop-cc-gui review

Desktop CC GUI is a bilingual, English-and-Chinese desktop client for AI coding tools. It puts Claude Code, Codex CLI, OpenCode, DeepSeek Harness, and 6 other engines in one Tauri interface, with chat, files, Git, a terminal, sessions, and model controls.

Verdict

Our desktop-cc-gui run installed 581 packages, built in 55 seconds, and passed all 1,231 tests in 110 seconds. That is the cleanest lab result in this batch and earns a trial for developers who regularly switch between Claude Code, Codex, and other CLIs. Keep the trial narrow until provider setup, shared-drive workspaces, plugins, and LAN access behave correctly on your machine.

We ran it

Lab card: what happened when we ran desktop-cc-guiScreenshot of desktop-cc-gui (www.mossx.ai/download)
Install✓ · 12s581 packages · 492 MB
Build✓ · 55s
Tests✓ · 110s1231 passed · 0 failed of 1231 (vitest)
Repo1199 files~199,273 lines of source · 14.1 MB · 3 CI workflows · tests dir

Answers from our run

Does desktop-cc-gui build from source?

Dependencies installed in 12 seconds (581 packages), and the build succeeded in 55 seconds. We cloned commit 0d042c3 into a clean Debian container with 3 CPUs and no project-specific setup.

Do desktop-cc-gui's tests pass?

Yes: 1231 of 1231 passed when we ran the project's own test command (vitest). Some failures need services or credentials a bare container does not have.

Who should not use desktop-cc-gui?

Buyers expecting bundled models or one account for every engine: each CLI and its provider access still have to be installed and configured.

What are the alternatives to desktop-cc-gui?

CloudCLI, Opcode, OpenCode. Our desktop-cc-gui run installed 581 packages, built in 55 seconds, and passed all 1,231 tests in 110 seconds.

Setup4/512-second install and green build; desktop prerequisites remain
Docs5/5English and Chinese guides cover engines, platforms, and internals
Community4/54,376 stars with active reports and pull requests on September 25
Maturity3/51,231 tests passed, but v1.0.9 has workflow-blocking reports

Who it’s for

Developers who already use 2 or more supported coding CLIs and want one graphical workspace.
Claude Code and Codex users who want visible tool calls, diffs, session history, and per-tab model controls.
Teams comfortable keeping provider credentials in each CLI's own configuration.
Tauri contributors who can work across React, TypeScript, Rust, platform packaging, and a plugin SDK.

Who it’s NOT for

Buyers expecting bundled models or one account for every engine: each CLI and its provider access still have to be installed and configured.
Teams requiring identical behavior across engines: the README matrix lists missing image input, question cards, reasoning controls, commands, and team mode depending on the CLI.
Windows users who keep projects on SMB shares without a trial first: open issue 1247 reports the app becoming unresponsive on one mounted workspace while local paths worked.
Anyone needing a calm release channel: v1.0.9 shipped with same-day reports that messages could not be submitted and adding a provider could crash the app.
Security teams unwilling to audit a desktop client that can run coding agents, a PTY terminal, plugins, Git actions, config writes, and a token-protected LAN bridge.

Setup reality

Our sandbox installed 581 pnpm packages in 12 seconds and used 492 MB. The build succeeded in 55 seconds, then all 1,231 Vitest tests passed in 110 seconds. The 14.1 MB checkout held 1,199 files and about 199,273 source lines.

Running the full desktop app adds Node.js 20+, pnpm 10, stable Rust, and Tauri's platform packages. Useful sessions also need at least one supported CLI plus that engine's account, API key, or provider configuration.

The repository is a workspace monorepo with 3 CI workflow files, a tests directory, and no Dockerfile. Installer builds differ by operating system; Linux needs WebKitGTK packages, Windows needs C++ Build Tools and WebView2, and macOS needs Apple's command-line tools.

Ten coding engines share one desktop with different limits

Desktop CC GUI 1.0.9 wraps 10 coding engines in a native Tauri application. The list includes Claude Code, Codex CLI, Kimi, Grok, Pi, OMP, DeepSeek Harness, Antigravity, OpenCode, and Qoder. A developer can open a project, choose an engine per conversation, watch tool calls, edit files, inspect Git changes, and use a PTY terminal without arranging separate terminal windows for each agent.

The common interface does not make the engines interchangeable. The README's compatibility table records missing or partial support for images, reasoning effort, question cards, slash commands, provider switching, and sub-agent displays. None of the 10 engines has team mode in the current matrix. This honesty is useful: choosing an engine still changes what a conversation can do, and ccgui cannot add a protocol feature that the underlying CLI does not expose.

All 1,231 Vitest checks passed in our sandbox

Our sandbox installed 581 pnpm packages in 12 seconds, taking 492 MB on disk. The production frontend build passed in 55 seconds. Vitest then completed 1,231 tests with 0 failures in 110 seconds. Among four measured repositories in this assignment, desktop-cc-gui is the only one whose install, build, and test commands all finished successfully, making its repository mechanics the easiest to trust on first contact.

The checkout at commit 0d042c3 contained 1,199 files, about 199,273 lines of source, and 14.1 MB before installation. It is a workspace monorepo with a separate plugin SDK, React frontend, and Rust backend. Our run measured the Node-side commands supplied by the lab. It did not launch a graphical Tauri window, authenticate paid model accounts, exercise every CLI adapter, or prove that a packaged installer works on all 3 operating systems.

The 492 MB install excludes Rust and platform packages

Source development requires Node.js 20 or newer, pnpm 10, stable Rust, and the operating system packages Tauri needs. Linux adds WebKitGTK libraries, Windows adds Microsoft C++ Build Tools and WebView2, and macOS needs Apple's command-line tools. The README warns that a first development launch compiles the Rust backend and may take several minutes. That work sits outside our measured 55-second frontend build.

A useful setup also needs at least 1 external coding CLI and its provider access. Ccgui writes provider channels to the native configuration files used by each CLI instead of keeping another credential database. That design avoids copying every secret into a new store, but the desktop app can still launch those tools and edit their configuration. Teams should start with one disposable project and one provider before importing a full set of accounts.

Version 1.0.9 fixed Codex effort and drew 2 blocking reports

Release v1.0.9 landed on September 24, 2026. Its change list includes a fix for Codex reasoning effort falling back to medium, worktree sub-workspaces, task details, provider filtering, and plugin loading repairs. Pull request 1293 says the release workflow initially hit a Windows Rust test failure caused by path escaping and skipped the Windows installer step; the merged fix also carried the Codex effort change back to main.

Two reports filed around that release touch basic interaction. Issue 1300 says Enter would not submit a message and the send arrow stayed disabled. Issue 1297 contains a v1.0.9 crash while adding a provider on Windows 10. Open issue 1247 describes an unresponsive workspace on one Windows 11 SMB path even though other shares and local paths worked. These reports are narrow, but each can block normal use before model quality matters.

Three CI workflows support an unusually active codebase

GitHub showed 3 CI workflow files, 4,376 stars, and 308 open issues and pull requests on September 25, 2026. The last push was September 24, the date of v1.0.9. Recent pull requests contain detailed failure reproductions, baseline comparisons, and focused tests across both languages. The 1,231 passing tests in our run reinforce that activity, while the 308-item combined queue warns that buyers are adopting a fast-moving application rather than a settled utility.

Documentation is a strength. English is the default README, a Chinese version is linked beside it, and the app ships both interface languages. Setup tables name platform dependencies, the engine matrix states limitations, and the development section maps frontend and Rust responsibilities. Some deeper plugin and UI documents are Chinese-only, so English-speaking plugin authors may need translation even though users can install and operate the app from the English guide.

LAN access and plugins expand the security review

The application can expose its interface to LAN devices through a token-authenticated WebSocket bridge. It also has a plugin SDK, a runtime permission boundary, file editing, Git controls, a terminal, and permission modes that include bypass for some engines. Each feature is reasonable for an agent workstation. Together they create a large local authority surface, especially when the selected CLI can execute commands or write outside the current project.

Desktop CC GUI is most convincing as a personal cockpit for someone already juggling 2 or more agent CLIs. The 12-second install, green build, and 1,231 passing tests justify trying it. They do not settle provider crashes, network-share hangs, plugin trust, or whether the GUI preserves every engine's safety choices. Keep bypass modes off, leave LAN access disabled until needed, and compare a changed file set before committing agent work.

Alternatives

ProjectWhat it isPick it when
CloudCLIA web and mobile interface for Claude Code, OpenCode, Cursor CLI, and Codex sessions.pick this instead when remote browser access matters more than a native Tauri desktop workspace.
Opcode gh↗A desktop GUI centered on Claude Code sessions, custom agents, and background work.pick this instead when Claude Code is your main engine and a narrower interface is preferable.
OpenCode gh↗A coding agent with its own terminal interface and provider support.pick this instead when one agent and a terminal-first workflow are enough.

What people are saying

  1. [github-trending] zhukunpenglinyutong/desktop-cc-gui

Sources

  1. Desktop CC GUI repository and English README
  2. Desktop CC GUI v1.0.9 release
  3. Issue 1300: message cannot be submitted
  4. Issue 1297: provider setup crash in v1.0.9
  5. Issue 1247: unresponsive SMB workspace
  6. Pull request 1293: Windows release and Codex effort fixes

More ai tools reviews

nobodywho · langextract · openrig · skills · awesome-grokbot · monty · the whole board →