Eliminate IDEA‑AI Switching with CC GUI: The Seamless Claude Code & Codex Companion

CC GUI – IDE‑integrated workbench for Claude Code and Codex

CC GUI (formerly Claude Code GUI) is an MIT‑licensed, fully open‑source JetBrains plugin (≈3.5 k GitHub stars). It embeds the capabilities of Claude Code and Codex CLI tools into a single IDEA panel, removing the need to switch between the IDE and external terminals.

Core capabilities

Instantly switch between Claude Code and Codex without installing separate plugins.

Use @file to attach files or images, providing richer context than raw CLI prompts.

Diff results appear directly in IDEA’s Diff view with clickable links to the affected lines.

History search, favorites, export, and token‑cost statistics show where AI budget is spent.

Built‑in slash commands and optional MCP server extensions add extra functionality.

Difference from official ACP

Core positioning: ACP runs an Agent inside the IDE; CC GUI provides a complete visual workbench.

Key capabilities: ACP focuses on lightweight diff view and context sharing; CC GUI adds session management, image input, Skill system, and MCP integration.

Target audience: ACP suits heavy CLI users; CC GUI targets developers who want to finish tasks entirely inside the IDE.

Trade‑offs: ACP is lightweight and CLI‑centric; CC GUI is richer and GUI‑centric.

The two approaches are not mutually exclusive; both can be installed and switched based on the scenario.

Installation (three steps)

Install the plugin. Open Settings → Plugins, search for “CC GUI”, and install the version with the highest download count (≈385.6 k downloads). The right‑side tool panel appears after installation.

Configure a provider. Choose one of four methods:

Create an API key in the Anthropic Console (separate from Claude.ai Pro/Max subscription).

Import a local ~/.claude/settings.json file that already contains an endpoint.

Use cc-switch, a community‑wide provider manager.

Specify a custom third‑party proxy endpoint (convenient for users in China).

Start a conversation. Use the right‑side panel to ask a simple question, e.g., “Analyze the current project structure”.

/plan mode – architecture‑first workflow

In high‑risk scenarios (e.g., payment integration), directly modifying code can be disastrous. /plan forces the AI to act as an architect, producing a detailed plan without touching the code. /plan 设计下银联的接入方案 The AI scans the project, reads PaymentService, extracts key points from the payment documentation, and returns a structured plan that includes:

Files that need modification.

New classes to add.

Required SDK dependencies.

Where signatures and encryption should be applied.

Callback and asynchronous notification handling.

After reviewing and approving the plan, switch to /agent mode to let the AI apply the changes.

Token‑cost statistics

Simple conversation (5 rounds): ~15 K tokens.

Code review (≈100 lines): ~25 K tokens. /plan for a medium‑size project: ~40 K tokens.

These numbers are displayed in the built‑in usage panel and are converted to approximate USD cost.

AI Commit and AI Review

AI Commit generates Conventional Commit messages from staged diffs, speeding up commit creation by about five times. AI Review runs the AI over a diff before a commit or PR, flagging boundary conditions, null‑pointer risks, potential infinite loops, and naming inconsistencies.

MCP server configuration

Enter the server address, headers, and authentication token once; the same configuration is shared with Claude Desktop and Cursor. Installing the Chrome DevTools MCP extension lets the agent invoke Chrome for front‑end testing that requires a logged‑in session.

Comparison with Qoder and CodeBuddy

Nature: CC GUI is a GUI wrapper for CLI agents; Qoder and CodeBuddy are standalone AI agents.

Open‑source: CC GUI is MIT‑licensed; the others are closed‑source.

Model support: CC GUI supports Claude, Codex, and custom models; Qoder and CodeBuddy use built‑in models.

Entry barrier: CC GUI requires a Claude/Codex subscription; Qoder offers a free tier; CodeBuddy provides a free personal edition.

Strengths: CC GUI reuses existing subscriptions and benefits from community contributions; Qoder is deeply optimized for the Java ecosystem; CodeBuddy integrates tightly with Tencent’s ecosystem.

Choosing logic: if you already have Claude Code or Codex, pick CC GUI to reuse your subscription. If you prefer a zero‑cost trial, try Qoder. If you are in the Tencent cloud environment, CodeBuddy may be the best fit. All three can be installed simultaneously without conflict.

Limitations

The quality of AI output depends on the accuracy of the supplied context; misuse of @file yields garbage‑in‑garbage‑out.

Clear definition of project boundaries is required; the AI cannot infer business constraints on its own.

Reviewing AI‑generated code is essential—code generation is easy, code review is hard.

Recommended starter tasks

Adding unit tests.

Generating boilerplate code.

Outlining module logic.

Converting screenshots into initial code snippets.

Project repository

https://github.com/zhukunpenglinyutong/jetbrains-cc-gui